t!
ill
11 I
"111
Vbr
:..:
Hi
mill
iiiiiijiii
I
BRAR^
'>i
in
University of California.
<
VI
OK
Class
ON THE
ALGEBRAICAL AND NUMERICAL
THEORY OF ERRORS OF OBSERVATIONS
AND THE
COMBINATION OF OBSERVATIONS.
ON THE
ALGEBRAICAL AND NUMERICAL
THEORY
OF
ERRORS OF OBSERVATIONS
AND THE
COMBINATION OF OBSERVATIONS.
By SIR GEORGE BIDDELL AIRY,
ASTRONOMER ROYAL.
SECOND
'
EDITION, REVISED.
Hon&on
aIACMILLAN AND
1875.
[All Eights
reserved.]
CO.
K.C.B.
QA
Camfcrt'trge
PRINTED BY
C. J.
CLAY.
MA
AT THE UNIVERSITY PRESS
PEEFACE TO THE FIEST EDITION.
The Theory
of Probabilities
divided into two parts.
is
naturally and strongly
One of these
relates to those
chances which can be altered only by the changes of
entire units or integral multiples of units in the fundamental conditions of the problem as in the instances
;
number of dots exhibited by the upper surface
of a die, or the numbers of black and white balls to
be extracted from a bag. The other relates to those
of the
chances which have respect to insensible gradations in
as in the duration
of life, or in the amount of error incident to an astrothe value of the element measured
nomical observation.
Tt may be difficult to commence the investigations
proper for the second division of the theory without
referring to principles derived from the first. Neverit is certain that, when the elements of the
second division of the theory are established, all reference to the first division is laid aside and the original
connexion is, by the great majority of persons who use
theless,
The two
the second division, entirely forgotten.
sions branch off into totally unconnected subjects
persons
who habitually use one
for the other
and
divi
those
part never have occasion
practically they
become two
different
sciences.
In order to spare astronomers
and observers
in
natural philosophy the confusion and loss of time which
are produced by referring to the ordinary treatises em
bracing both branches of Probabilities,
have thought
PREFACE.
VI
it
desirable to
draw up
this tract, relating only to Errors
of Observation, and to the rules, derivable from the
consideration of these Errors, for the Combination of
the Results of Observations.
have
thus also the
advantage of entering somewhat more fully into several
points, of interest to the observer, than can possibly be
done
in a General
Xo
Theory of Probabilities.
novelty, I believe, of fundamental character, will
be found in these pages. At the same time I may state
work has been written without reference to
or distinct recollection of any other treatise (excepting
only Laplace's Theorie des Probabilites) and the methods of treating the different problems may therefore
differ in some small degrees from those commonly emthat the
ployed.
G. B.
AIRY.
Royal Obseevatoey, Greenwich,
January 12, 1861.
PREFACE TO THE SECOND EDITION.
The work has been thoroughly revised, but no important alteration has been made except in the introduction of the new Section 15, and the consequent
alteration in the numeration of articles of Sections 1(5
:
and 17 (formerly 15 and 10) and in the addition of the
Appendix, giving the result of a comparison of the
theoretical law of Frequency of Errors with the Frequency actually observed in an extensive series.
:
G. B.
February
20,
1875.
AIRY.
INDEX.
PART
I.
FALLIBLE MEASURES, AND SIMPLE ERRORS OF OBSERVATION.
Section
1.
Nature of the Errors here considered.
PAGE
Article
2.
Instance of Errors of Integers
3.
Instance of Graduated Errors
these are the sub
ject of this Treatise
....
4.
Errors of an intermediate class
5.
Instances of Mistakes
C.
Characteristics of the Errors considered in this
Section
i).
The word Error really means Uncertainty
2.
ib.
ib.
Treatise
8.
Laic of Probability of Errors of any given
amount.
Reference to ordinary theory of Chances
10.
Illustrations of the nature of the law
11.
Illusfration of the algebraic
for the
12.
Laplace's investigation introduced
13.
Algebraical
causes of error assumed
of
...
ib.
form to be expected
law
combination
...
6
7
many independent
ib.
INDEX.
Vlll
PAGE
Article 15.
This leads to a definite integral
16.
Simplification of the integral
17.
Investigation of
J
S
10
11
dt.e~&
Jo
IS.
Investigation of
f"
I
.12
'
dt. cos rt. e~ 1
2
.
Jo
20.
Probability that an error will
x + 8x is found to be
between x and
fall
a 2
t .e~c 2
.8x
.14
cJtt
.15
21.
Other suppositions lead to the same
22.
Plausibility of this law
23.
Curve representing the law of Frequency of Error
result
table of values of
ib>
16
3.
Consequences of the Law of Probability or
Frequency of Errors, as applied to One System of
Measures of One Element.
Section
25.
It is
assumed that the law
of Probability applies
equally to positive and to negative errors
"Mean Error"
26.
Investigation of
27.
Investigation of "Error of
28.
Definition of "Probable Error"
29.
Tableof/ dic.~ K2 and
Mean Square"
.
.IS
.19
.
investigation of Pro
Jo
bable Error
30.
Remark on
the small
22
number
of errors of large
value
31.
32.
20
.21
/""'
7''
's/
23
Table exhibiting the relations of the Modulus and
the several Errors
ib.
Introduction of the term "Actual Error"
24
INDEX.
Section
Remarks on
4.
the application
of
these processes
in particular cases.
PAGE
"With a limited
Article 33.
number
of errors, the laws will be
imperfectly followed
24
Case of a single discordant observation
34.
PART
ERRORS IX
TIIE
.25
II.
C0MBIXATI0X OP FALLIBLE MEASURES.
Law of Frequency of Error, and values
5.
Mean Error and Probable Error, of a symbolical
Section
of
or
numerical Multiple of One Fallible Measure.
The Law of Frequency has the same form as for
the original: the Modulus and the Mean and
35.
Probable Errors are increased in the proportion
expressed by the Multiple
36.
The multiple
37.
Nor the sum
itself a
of measure here considered
is
simple measure
of
26
not
27
numerous independent measures
ih.
Law
of Frequency of Error, and values of
Error, of a quantity formed
by the algebraical sum, or difference of two independent
Fallible Measures.
Sectiox
6.
Mean Error and Probable
39.
The problem
is
reduced to the form of sums of
groups of Errors, the magnitudes of the errors
through each group being equal.
.29
.
43.
Results
that, for
the
Fallible Measures, the
sum
of two independent
Law
of Frequency has the
same form as for the originals, but the square of
the new modulus is equal to the sum of the
squares of the two original moduli
.33
INDEX.
PAGE
The same theorem of magnitudes applies to Mean
Error, Error of Mean Square, and Probable
Article 44.
Error
33
45.
But the combined
47.
The same formulae apply
Measures must be ab
Fallible
solutely independent
34
for the difference of
independent Fallible Measures
two
.30
all cases here to be treated, the Law of Frequency has the same form as for original obser
In
49.
vations
Section*
37
Values of Mean Error and Probable Error
combinations which occur most frequently.
7.
in
50.
Probable Error of kX+l If
51.
Probable Error of
52.
Probable Error of rIZ + sS+tT+uU+&.c.
53.
Probable Error of
3s
R + S+T+U+&C
X +A
ib.
33
where
2 + ...4X,
the
quantities are independent but have equal probable errors
54.
bable error of
55.
Section
5G.
(b.
Difference between this result and that for the pro
nXi
Probable Error of the Mean of
8.
40
X X
1
i} ...X
41
Instances of the application of these Theorems.
Determination of geographical eolatitude by observations of zenith distances of a star above ami
below the pole
57.
42
Determination of geographical longitude by transits of
the
Moon
43
INDEX.
XI
Section 9. Methods of determining Mean Error and
Probable Error in a gicen series of observations.
PAGE
Article 58.
The
peculiarity of the case
the quantity measured
is
is,
that the real value of
not certainly
known
59,
For the Mean Error, the rule
60,
For Error of Mean Square, and Probable Error,
is
before
the divisor of
of
61,
44
the same as
ib.
sum
n 1
of squares will be
instead
45
Convenient methods of forming the requisite numbers
47
FART
III.
PRINCIPLES OF FORMING THE MOST ADVANTAGEOUS COMBINA
TION OF FALLIBLE MEASURES.
Section 10. Method of combining measures ; meaning of
" combinationweight ;" principle of most advantageous
combination ; caution in
its
application to "entangled
measures."
62.
First class of
measures
quantity which
are
direct
is invariable,
measures of a
or whose variations
known
49
63.
Combination by means of combinationweights
64.
The combination
to
be sought
is
that which will
give a result whose probable error
is
the smallest
possible
ib.
65.
To be found by the algebraical theory
maxima and minima
6G.
Sometimes, even for a simple
occur " entangled measures."
reduction of these
50
result,
of complex
51
there will
Caution fur the
ib.
XU
INDEX.
PAGE
Article 67.
Second
measures
class of
when
the corrections to
several physical elements are to he determined
simultaneously
this is also
....
a case of algebraical
complex maxima and minima
52
11.
Combination of simple measures; meaning
of " theoretical weight;" simplicity of results for theoretical weight; allowable departure from the strict
Section
rules.
68.
Independent measures or results are supposed
equally good the investigation shews that they
must be combined with equal weights
;
69.
53
Independent measures
good
their combinationweights must be
or results are not equally
in
versely proportional to the square of the probable
error of each
70.
54
If the reciprocal
of
(probable error)
"theoretical weight,"
their
be
called
combinationweight
be proportional to their theoretical
and the theoretical Aveight of result
= sum of theoretical weights of original mea
oucrht
weight
to
;
sures
">.")
72.
Instance
73.
We may depart
eolatitude by different stars
somewhat from the
strict rules for
formation of combinationweights without introducing material error of result
,
Section
74.
12.
Instance
transits
Longitude is determined by lunar
compared with those at two known
(1).
60
Reference must be made to actual errors
for combinationweights,
of result
58
Treatment of entangled measures.
stations
75.
"><;
and
for theoretical
result
weight
(jl
INDEX.
Xlll
PAGE
Article 76.
Partition of theoretical weight of result
78.
Partition
79.
Instance
is
applicable in other cases
(2).
....
theoretical weight for
each azimuth found by partition
ib.
Zenith distances of stars are observed at three stations of a meridional arc; to
Instance
(3).
find the amplitude of the first section
81.
.62
.04
Theodolite observations of the meri
dian and of distant signals
SO.
ib.
All valid combinations must be considered, and,
being entangled observations, must be treated by
.65
actual errors
82.
Equations formed and solved
84.
The
06
and that for the
second section are entangled, and cannot be comresult for the first section
bined to form the result for the whole
difference
between actual error and probable error
86.
vations
Section
to
"
13.
several
70
Treatment of numerous equations applying
quantities; introduction of the term
unknown
minimum
squares.'"
.71
87.
General form of such equations
88.
Obvious method of combining them in order to
form the proper number of determining equa
tions
89.
72
Symbolical equations for x, one of the unknown
quantities
90.
Symbolical
error of
91.
6S
General caution for treatment of entangled obser
73
equations
for
making the probable
x minimum
Synthetical solution of the equations
74
.
.75
INDEX.
XIV
PAGE
Article 93.
94.
Complete exhibition of the form of solution
.77
is the same as the form of solution of
the problem, " to reduce to minimum the sum of
This form
when
squares of residual errors,
"
minimum
the errors are
Introduction of the term
properly multiplied."
Danger of using
squares."
96.
Expression for probable error of
97.
Approximate values of the
term
this
78
.79
factors will suffice in
practice
Section
14.
SI
Instances of the formation of equations ap
plying
99.
Instance
tions
100.
to
1.
quantities.
Determination of the personal equa
among
Instance
unknown
several
several transitobservers
.82
Consideration of a net of geodetic
2.
triangles
101.
So
The probable error
ascertained
of each measure
different
for angles
must first be
between sta
tions, for absolute azimuths, for linear
measures
ib.
102.
Approximate numerical coordinates of stations are
to be assumed, with symbols for corrections
.
S6
103.
Corresponding equations for measures mentioned
above
ih.
104.
These equations
SS
105.
Generality and beauty of the theory
10G.
No
will suffice
;
it
application to any supposed measures
objection, that the
admits of
instance
measures are heteroge
neous
107.
Solution of equations
S9
ih
is
troublesome
.90
INDEX.
XV
Treatment of Observations in which it is
15.
required that the Errors of Observations rigorously
Section
satisfy
some assigned conditions.
PAGE
Article 110.
In a geodetic triangle, of which the
three angles are observed, and their sum proves
erroneous to find the corrections for the several
Instance
1.
angles
91
111.
Equations for probable errors
112.
Assigned condition introduced
113.
Result
114.
Instance
....
92
ib.
2.
In a series of successive azimuthal
whose sum ought to be 360, the sum
angles,
proves erroneous
find
to
the corrections for
the several angles
94
115.
Result
116.
Instance
117.
Assigned conditions introduced
ib.
3.
station, all
In a geodetic hexagon, with a central
the angles are subject to error
....
.
Eliminations, and equations
118,119,120.
122.
ib.
Practical process, which
PART
may be
preferable
95
96
97,99,100
.101
IV.
OX MIXED ERRORS OF DIFFERENT CLASSES, AND CONSTANT
ERRORS.
Section
16.
Consideration of the circumstances under
of Mixed Errors of Different
which the existence
Classes
may
be recognized ;
and
investigation of their
separate values.
124.
The existence of Error of a Different Class
to be assumed without good evidence
.
125.
is
.
not
.103
Especially without evidence of possibility of such
Error
ib.
INDEX.
XVI
Articlel26.
....
Formation of result of each group
127.
Discordance of results of different groups
12S.
Investigation of
129.
Decision on the reality of a
130.
Much must depend on
Mean
li>4
ib.
Discordance, supposed to
be a mutter of chance, and
its
Probable Error
Mean Discordance
the judgment of the
105
ib.
Com
puter
131.
PAGE
106
Simpler treatment when Discordance appears to
be connected with an assignable cause
ib.
Section 17. Treatment of observations when the values
of Probable Constant Error for different groups, and
probable error of observation of individual measures
within each group, are assumed as known.
132.
"We must not in general assume a value for Constant Error for each group, but must treat it
133.
Symbolical formation of actual errors
134.
Symbolical formation of probable error of result
135.
Resulting combinationweights
136.
Simpler treatment when the existence of a definite
Constant Error for one group is assumed
.Ill
as a chanceerror
equations of
107
.
minimum
10S
109
.
.110
CONCLUSION.
137.
Indication of the principal sources of error and
inconvenience,
in the applications which have
been made of the Theory of Errors of Observations and of the Combination of Observations
.
112
APPENDIX.
Practical Verification of the Theoretical
quency of Errors
Law
for the Fre
.114
COEKIGENDA.
Page
47, dele line 1,
and
Mean Square
Page
61,
of
substitute the following:
Sum
of Errors
between lines 6 and
"
affected by the
7,
a + b + c + d + &e.
"final apparent results, as
insert
actual error of
read
,,
line 12, for
line 14, for
'
,,
actual errors of the
line 19, for
'
,,
actual error
'
'
'
read
'
'
apparent
read
'
result
'.
'
'.
apparent
'.
OF THE
UNIVERSn
ON THE
ALGEBRAICAL AND NUMERICAL THEORY
OP
ERRORS OF OBSERVATIONS
AND THE
COMBINATION OF OBSERVATIONS.
PART
I.
AND SIMPLE ERRORS OF
FALLIBLE MEASURES,
OBSERVATION.
1.
The
1.
Nature of
the
Errors here considered.
nature of the Errors of Observation which
form the subject of the following Treatise, will perhaps
be understood from a comparison of the different kinds
of Errors to
which
different Estimations or
Measures are
liable.
Suppose that a quantity of common nuts are put
and a person makes an estimate of the num
2.
into a cup,
ber.
will
His estimate may be correct
be
A.
incorrect.
But
if
incorrect,
more probably
it
the error has this
SIMPLE ERRORS OF OBSERVATION.
peculiarity, that
it is
may be
to
This class of errors
These are not the errors
called Errors of Integers.
which
There can
an error of whole nuts.
not be an error of a fraction of a nut.
this treatise applies.
Instead of nuts, suppose water to be put into the
3.
cup, and suppose an estimate of the quantity of water to
be formed, expressed either by
its
weight.
its
cubical content, or
Either of those estimates
any amount
(practically not
may be
in error
exceeding a certain
by
by
limit),
proceeding by any gradations of magnitude, however mi
This class of errors
nute.
It is to the
Errors.
may
be called Graduated
consideration of these errors that
this treatise is directed.
4.
If,
instead of nuts or water, the cup be charged
with particles of very small dimensions, as grains of fine
sand, the state of things will be intermediate between the
two considered above.
Theoretically, the errors of esti
mation, however expressed, must be Errors of Integers of
SandGrains
so small that
but practically, these sandgrains
it
is
may be
a matter of indifference whether the
gradations of error proceed
by whole sandgrains
fractions of a sandgrain.
In this
practically
5.
kind
In
case,
or
the errors are
Graduated Errors.
all
these cases, the estimation
is
of a simple
but there are other cases in which the process
be either simple or complex and, if it is complex, a
Suppose,
ferent class of errors may be introduced.
;
instance,
by
it is
desired to
know
may
dif
for
the length of a given road.
NATURE OF THE ERRORS.
may
person accustomed to roadmeasures
length
this estimation will
and
this
method
estimate
its
be subject simply to Graduated
Another person may measure
Errors.
measure
its
length by a yard
of measuring, from uncertainties
in the adjustments of the successive yards, &c. will also
subject to Graduated Errors.
But besides
be
this, it will
be
subject to the possibility of the omission of registry of
entire yards, or the record of too
many
entire yards
not
as a fault of estimate, but as a result of mental confusion.
/
In like manner, when a measure
meter
there
may
is
made with
represented by the fractional part of the reading
may
also be error of the
of the whole
a micro
be inaccuracy in the observation as
number
number
but there
of whole revolutions, or
of decades of subdivisions, similar to
the erroneous records of yards mentioned above, arising
from causes totally distinct from those which produce
accuracy of mere observation.
be called Mistakes.
This class of Errors
Their distinguishing peculiarity
that they admit of Conjectural Correction.
in
may
is,
These Mistakes
are not further considered in the present treatise.
G.
The
errors therefore, to
vestigations apply,
may be
the following conditions
which the subsequent
in
considered as characterized by
They are infinitesimally graduated,
They do not admit of conjectural correction.
7.
will
Observations or measures subject to these errors
be called in this treatise "fallible observations," or
" fallible measures."
A2
SIMPLE ERRORS OF OBSERVATION.
we
Strictly speaking,
8.
ought, in the expression of
our general idea, to use the word "uncertainty" instead of
For we cannot at any time
"error."
that our estimate or measure, though fallible,
fectly correct
no "
and therefore
it
may happen
error," in the ordinary sense of
positively
assert
is
not per
that there
And,
the word.
is
in
manner, when from the general or abstract idea we
like
proceed to concrete numerical evaluations, we ought, instead
of "error," to say "uncertain error;" including,
among
the uncertainties of value, the possible case that the un
may = 0. With
certain error
may
When
still
be
the term
is
"Actual Error" or "Probable Error," there
is
used without danger of incorrectness.
qualified, as
the
this caution, however, in
interpretation of our word, the term " error "
no fear of misinterpretation.
Law
2.
of Probability of Errors of any given amount.
In estimating numerically the "probability" that
9.
the magnitude of an error will be included between two
given limits, we shall adopt the same principle as in the
ordinary Theory of Chances.
of the "probability"
shall consider
all
is
When
to be
the numerical value
determined a priori, we
the possible combinations which pro
duce error; and the
fraction,
whose numerator
is
ber of combinations producing an error which
between the given
total
number
limits,
the
is
num
included
and whose denominator
is
the
of possible combinations, will be the "pro
bability
"
limits.
But when the numerical value
that the error will be included between those
is
to
be deter
LAW OF PROBABILITY OF ERRORS.
mined from observations, then if the numerator be the
number of observations, whose errors fall within the given
limits, and if the denominator be the total number of
observations, the fraction so formed, when the number of
observations
10.
is
indefinitely great,
First, that,
two conclusions
though there
tween the two values
small
is
in
is,
any given
case, a pos
and therefore a
magnitude of an error may
possibility that the
fall
" probability."
of a large magnitude,
sibility of errors
still it
the
very slight contemplation of the nature of
errors will lead us to
may
is
and
E + he,
where
fall
is
be
large
more probable that the magnitude of an error
e and e + he, where e is
between the two values
he being
supposed to be the same in both.
in estimating the length of a road,
the estimator's error will
101 yards than that
11 yards.
Or,
if
it
fall
Thus,
probable that
between 100 yards and
between 10 yards and
will fall
the distance
it is less
is
measured with a yard
measure, and mistakes are put out of consideration,
less likely that
the error will
101 inches than that
it
it is
between 100 inches and
fall
will fall
between 10 inches and
11 inches.
Second, that, according to the accuracy of the methods
used and the care bestowed upon them, different values
must be assumed
for
the errors in order to present com
parable degrees of probability.
Thus, in estimating the
roadlengths by eye, an error amounting to 10 yards
sufficiently probable
may fall between 10
and the chance that the
yards and 11 yards
is
is
real error
not contemptibly
SIMPLE ERRORS OF OBSERVATION.
But
small.
ficant that
in measuring
by a yardmeasure, the proba
the error can amount to 10 yards
bility that
no
man
think
will
and 11 yards
will
so insigni
worth consideration
it
the probability that the error
is
may
and
between 10 yards
fall
never enter into our thoughts.
It
may,
however, perhaps be judged that an error amounting to 10
inches
about as probable with this kind of measure as
is
an error of 10 yards with eyeestimation
lity
that the error may
mode
with this
and the probabi
between 10 inches and 11 inches,
fall
of measuring,
may
be comparable with the
probability of the error, in the rougher estimation, falling
between 10 yards and 11 yards.
Here then we are led
11.
braical formula
error will fall
which
to the idea that the alge
to express the probability that
is
between the limits
and
+ Be
extremely small) will possess the following properties
Inasmuch
(A)
interval of limits,
The term
(B)
The term
or parameter
probabilities
sure,
and
measure.
must be
is
c,
of the form
when
cf>
(e)
same
Be.
as e increases,
and
e is indefinitely large.
must contain a constant symbol
< (e)
which
probability in the
must diminish
<p (e)
indefinitely small
(C)
is
by multiptying our very narrow
as,
we multiply our
proportion, the formula
must be
an
(where Be
constant in the expression of the
is
under the same system of estimation or mea
different for different systems of estimation or
If (as
seems
likely),
upon taking a proper pro
portion of magnitudes of error, the law of declension of
the probability of errors
is
the same for delicate measures
laplace's investigation of their law.
and
then the formula
for coarse measures,
form
yjr
() x 8 ()
or
i//
(J
where
will
be of the
c is small for
a delicate system of measures, and large for a coarse
system of measures.
[The reader
recommended,
is
in the first instance, to
pass over the articles 12 to 21.]
Laplace has investigated, by an a priori process,
12.
well worthy of that great mathematician, the form of the
Without enter
function expressing the law of probability.
ing into
all details, for
which we must refer to the Theorie
Analytique des Probabilites,
we may
give an idea here of
the principal steps of the process.
The fundamental
13.
principle in this investigation
that an error, as actually occurring in observation,
simple origin, but
tion of a great
is
may
many independent
causes of error
chance which
produce an
different magnitude.
not of
produced by the algebraical combina
which, according to the
pendently,
is
is,
error,
affects
of either sign
each of
it
inde
and of
These errors are supposed to be of
the class of Errors of Integers, which admit of being
by the usual Theory
treated
of
Chances
then, supposing
the integers to be indefinitely small, and the range of their
number
to
be indefinitely great, the conditions ultimately
approach to the state of Graduated Errors.
1
This
is
not the language of Laplace, but
standing on which his investigation
trrors of
observation.
is
most
it
appears to bo the iinder
distinctly applicable to single
SIMPLE ERRORS OF OBSERVATION.
Suppose then
14.
errors
may
, 7i +
be,
l,
one source of
that, for
?i
2,...l,
+1,
0,
the probability of each will be ^
Suppose that,
may
also be,
n,
and
on
Ti2,
2, ...
n1,
for another source of error, the errors
ra+2, ...1,
0,
1, 2,
...n2, n 1,
And
for s sources of error.
wish to ascertain what
suppose that
with one error taken from the second
and
error taken from the third series,
duce an error
The
I.
if
we
first
step
is,
series,
first series,
and with one
so on,
we can prohow many
to ascertain
are the different combinations which will each produce
Now,
n,
the probability that, upon com
is
bining algebraically one error taken from the
15.
with equal probability,
n + 1,
so
the
error,
with equal probability,
we watch
the process of combination,
I.
we
numbers are added by exactly the same
shall see that the
law as the addition of indices in the successive multiplications of the polynomial
n0 v'l
by
g(nl) e
therefore the
be, the coefficient
coefficient of e~
f
~n0 Vl
g(n2) fl
V"l
('i2)
V~l
(nl)
supposing the operation repeated
itself,
And
Vl
),
_i_
>H
times.
number of combinations required will
of e wvM (which is also the same as the
WvM
[nl)0 Vl
\M
in the expansion of
g1*8)0\'l
(2!0
V"l
^"1)0^1
^0 v'l
laplace's investigation of their law.
number uncom
This coefficient will be exhibited as a
bined with any power of
by
either
mV_1 or
,
The number
(6
e~
^" 1
Wyht or
,
same
we multiply the expansion
by \
(e
of combinations required
wVi__ z0Vi\
or the
if
term independent of
as the
by
Vi
as the
__
16
is
^ + e' ^
1
16
).
therefore the
same
in the expansion of
(D
Vi _[_
& c< + e !" 9v i^. e0Vii.s
/
term independent of
in the ex
pansion of
cos 10 x {1
+ 2 cos + 2
And, remarking that
respect to
0,
from
if
to
will entirely disappear,
be multiplied by
will
combinations required
\"d6 cos 10 x
.
cos n0} s .
integrate this quantity with
= ir,
the terms depending on
and the term independent of
it
follows that the
number
of
the definite integral
+ 2 cos 6 + 2 cos 20
...
+ 2 cos n0}
s
,
[*
7T
c70
cos 10
jo
And
are,
ir,
we
+2
Jo
7T
or
(1
is
cos 20
the total
number
of possible combinations which
priori, equally probable,
is
(2n
+ l) s
SIMPLE ERRORS OF OBSERVATION.
10
Consequently, the probability that the algebraical combination of errors, one taken from each series, will produce
the error
is
1,
[2n
+ l)*'7r
2n+l
/sin
f*
dd cos
.
19
s
2
In subsequent
n and s are supposed to be very
steps,
large.
To
16.
which
is
integrate this, with the kind of approximation
proper for the circumstances of the case, Laplace
assumes
2>i+l
sin
e
(2?i
(as
the exponential
+
is
1)
sin 
essentially positive, this does not in
strictness apply further
than
2n +
= it;
but as succeed
ing values of the fraction are small, and are raised to the
high power
with the
s,
first
powers of
6,
they
may be
safely neglected in comparison
part of the integral)
and the exponential
expanding the sines in
in powers of
it
will
be found that
AJ{n{n+ l)s}\
where
B is
a function of n which approaches, as n becomes
laplace's investigation of their law.
11
The
expres
very large, to the definite numerical value T^.
sion to be integrated then becomes,
VG
+ 1)
7T V{H('tt
'
S\
m
i,
a.cos
r77,^v<3 ,
_,.
l_Vi0Hl)sj
To simplify
fr+<xc.
H b
f,
V
this integral,
it is
'2
.(l+ +&C
2
be remarked that
to
e~ c
multiplies the whole, and that this factor decreases with
extreme rapidity as
terms
and when
in the
t is
While
increases.
small, the
is
argument of the cosine are unimportant;
large, it matters not
whether they are retained
or not, because their rejection merely produces a different
length of period for the periodical term which
by an excessively small
plied
Also
coefficient.
is
it
multi
appears
be shewn in Article 19) that the integration of
(as will
such a term as cos mt .e f\ oBt" introduces no infinite term,
and therefore when
s,
this
may be
it is
divided by the very large
The
rejected.
integral
is
number
therefore reduced
to this,
a
\/G
7r
'
As the
17.
dt
\j\n (n
e~ \
first
There
It
at cos
1) s] J
is
'
\/{n (n
\/G
'
1) s}
step to this, let us find the value "of
no process
for this
purpose so con
venient as the indirect one of ascertaining the solid content
of the solid of revolution in
which
is
the radius of any
SIMPLE ERRORS OF OBSERVATION.
12
= e" f\
Let # and
be the other rectangular coordinates, so that
f = x* + y~.
and
section,
y
Then
z the corresponding ordinate
may
the solid content
be expressed in either of the
following Yfays:
By
polar coordinates, solid content
= 2tT.
dt.t.~ P =7T.
Jo
By
rectangular coordinates, solid content
dx
dy e~^ + ^
.
oo
=(4 // e
x (/."'^
el\
=o
00
for
r^
?''
i'
what symbol
the independent variable.
Hence,
and
dt . e~^ )
18.
ac
i =i CC'
since, for a definite integral, it is indifferent
be used
cfy
"
dt
'
= 
Next, to find the value of
ir,
dt cos ri
.
e~ l \
Call
Jo
this definite integral y.
As
this is a function of r, it can
differentiated with respect to r;
and
bo
as the process of inte
gration expressed in the symbol does not apply to
r,
can be differentiated by differentiating under the integral
sign.
Thus
dt.t:sin rt.~\
^=f
dr
J o
laplace's investigation of their law.
Integrating by parts, the general integral for
=  sin rt
e~
'
in which, taking the integral
jdt
from
cos rt
term vanishes, and the second becomes
13
e~ fi,
to
oo
the
first
^ y. Thus we have
dy _
di
= ~ rv y
'
Integrating this differential equation in the ordinary
way,
y=
Now when r = 0, we
C.e'i.
have found by the
kiTT
the value
val
of y for that case
is

last article that
Hence we obtain
tin ally
at
19.
we
r, we
If
respect to
cos rt .e
differentiate
this
expression twice with
find,
dt.f. cos rt
e~ t2
^1
'I
and expressions
times, six times, &c.
never
infinite.
we
differentiate four
The righthand
expressions are
of similar character if
This
is
the theorem to which
we
referred
in Article 16, as justifying the rejection of certain terms in
the integral.
14
SIMPLE ERRORS OF OBSERVATION.
Reverting
20.
now
to the expression at the
and making the proper changes
Article 18,
find for the value of the integral at the
2yV' ij[n(n+
end of
of notation,
end of Article
we
1G,
e 4(n+l) t
l)s\
This expression for the probability that the error, pro
duced by the combination of numerous errors
be
14), will
of
I,
(see Article
based on the supposition that the changes
is
magnitude of
proceed by a unit at a time.
If
pass from Errors of Integers to Graduated Errors,
consider that
we have thus obtained
that the error will
obtain
all
and
between
lie
and
the probabilities
all
?
+ l.
In order to
the probabilities that the error will
lie
Ave derive the following expression
81,
now we
we may
between
from that
above,
fjir' \/{in
Here
Hsa
(n
+ 1)
e in(n+lj7a
s]
very large number, expressing the magni
tude x of an error which
is
not strikingly large, by a large
multiple of small units.
Let
= mx,
where
m is large
bability that the error falls
1
\/G
v%'
T
.Let
jl
/{4m
(n
l) .8
,/
"'
.
+ l).s}'

*"
e 4n.(+l).s
,
^ =
4n(n +
81= m8x and the probetween x and x + 8x is
;
c, wliere c
may
$x
be a quantity of
laplace's investigation of their law.
magnitude comparable
to the
15
magnitudes which we
shall
use in applications of the symbol x; then
we have
for the probability that the error will fall
between x and
x+
finally
Bx,
t
1
j
c
Bx.
C\fir
This function,
racters
We
be remarked, possesses the cha
it will
which in Article 11 we have indicated as necessary.
shall hereafter call c the modulus.
Laplace afterwards proceeds to consider the effect
21.
of supposing that the probabilities of individual errors, in
the different series mentioned in Article 14, are not uniform
through each
series, as is
supposed in Article 14, but vary
according to an algebraical law, giving equal probabilities
for
+ or errors
also
we
And
same magnitude.
of the
he finds a result of the same form.
For
in this case
this,
however,
refer to the Theorie Analytique des Probabilites.
Whatever may be thought
22.
which
this
formula has been obtained,
doubted by any one that the result
is
of the
it
process
will scarcely
In
errors.
order to exhibit the numerical law of frequency (that
gives a
c2
which,
number proportioned
between x and x
falling
puted
when
be
entirely in accord
ance with our general ideas of the frequency of
the variable factor
by
is,
multiplied by Bx,
to the probability of errors
Bx), the following table is
com
SIMPLE ERRORS OF OBSERVATION.
16
Table of Values of e~c\
LAW OF PROBABILITY OF
which the abscissa represents 
ERRORS.
17
or the proportion of the
magnitude of an error
to the modulus,
and the ordinate
represents the corresponding frequency of errors of that
magnitude.
Here
it
will
be remarked that the curve approaches
the abscissa by an almost uniform descent from Magnitude
of Error
to
Magnitude of Error
= 1*7
x Modulus
that after the Magnitude of Error amounts to
and
20 x Modulus,
the Frequency of Error becomes practically insensible. This
A.
SIMPLE ERRORS OF OBSERVATION.
18
is
precisely the kind of law
which we should a priori have
expected the Frequency of Error to follow; and which, without such an investigation as Laplace's,
assumed generally
and
we might have
which, having assumed
for
we might have searched an algebraical law.
reasons, we shall, through the rest of this treatise,
general form,
For these
assume the law of frequency
17 e
c y7r
'
c
Bx,
as expressing the probability of errors occurring with
nitude included between
3.
Consequences of the
mag
x and x + Bx.
Law
of Errors, as applied
of Probability or Frequency
One System of Measures of
to
One Element.
The Law
24.
Errors,
of Probability of Errors or
to this.
being a
Suppose the
total
number
very large number
errors,
Frequency of
which we have found, amounts practically
of Measures to be A,
may
then we
whose magnitudes
fall
expect the
between x and x
number
Bx, to
of
be
~
i e &
Bx,
C V7T
where
c is
a modulus, constant for
but different
for Different
One System
of Measures,
Systems of Measures.
It
is
partly
the object of the following investigations to give the means
of determining either the
related to
25.
it,
This
in
modulus
c,
or other constants
any given system of practical
may be
'
errors.
a convenient opportunity for remark
ing expressly that the fundamental suppositions of La
MEAN ERROR.
19
assume that the law of
place's investigation, Article 14,
Probability of Errors applies equally to positive and to
It follows therefore that the formula in
negative errors.
must be received as applying equally to positive
negative errors. The number A includes the whole
Article 24
and
to
may happen
of the measures, whether their errors
to be
positive or negative.
Conceive
26.
which
to be
is
sider the
now
that the true value of the Element
measured
is
known (we
more usual case when
it
shall hereafter con
not known), and
is
that the error of every individual measm^e can therefore
The
be found.
number which
the
mean
readiest
is
of inferring from these a
Modulus
closely related to the
to take
is,
and
of all the positive errors without sign,
mean
take the
method
to
of all the negative errors without sign
(which two means,
when the number
of observations
is
very great, ought not to differ sensibly), and to take the
numerical
mean
JVTean Error.
It
is
is
Since the
included between
to the
Its relation
number
of errors
x and x + Sx
is
be called the
mere numerical
to be regarded as a
quantity, without sign.
thus found.
may
This
of the two.
Modulus
is
whose magnitude
 x~
c2
Bx,
and the
c sjTT
magnitude of each error does not
the
the
sum
sum
differ sensibly
of these errors will be sensibly
J
r e
*
.
fi2
'
2
.
x,
xSx and
;
c sjir
of all the errors of positive sign will
A r,a x
from
_*!
be
cA
B 2
SIMPLE ERRORS OF OBSERVATION.
20
The number
of errors of positive sign
CtJlTjQ
Dividing the preceding expression by
Mean
is
error
positive
1
=
V 73
this,
"
Similarly,
Mean
And
negative error
therefore,
Mean
= =cx 0o64189.
Error
V7T
And
conversely,
c
By
= Mean
this formula, c can
of errors
27.
is
Error x 1 772454.
be found with ease when the series
exhibited.
It is
however sometimes convenient
(as will
ap
pear hereafter, Article 61) to use a method of deduction
derived from the Squares of Errors.
The
positive
and
negative errors are then included under the same formula.
we form the mean of the squares, and extract the square
mean, we may appropriately call it the Error
of Mean Square. This, like the Mean Error, is a numerical
quantity, without sign.
To investigate it in terms of c,
we remark that the sum of the squares of errors between
x and x + Bx (formed as in the last Article) w ill be
If
root of that
**
Bx X X2
ERROR OF MEAN SQUARE.
and the sum of
The
+0
all
the squares of errors will be
dx.e

and the second term
is .4.
<?
.x
+ co
Ac 2
= + ~
i
_*!
x.e
c2
and
oo
The whole number
Hence the Mean Square
Square
(Ac
i~
term vanishes between the limits
first
21
is
+x
of errors
and
,
the Error of
Mean
is
'1
cj^ =cxO707107;
or c
It
28.
= Error
of
Mean Square x T414214.
has however been customary to
make
different
number, called the Probable Error.
meant by
this
term that the number used
is
use of a
It is not
a more pro
bable value of error than any other value, but that,
the positive sign
errors larger
attached to
is
than that value
is
it,
number
the
about as great as the num
ber of positive errors smaller than that value
when the negative
sign
is
attached to
applies to the negative errors.
is
when
of positive
and
that,
the same remark
it,
The Probable Error itself
To ascertain the
a numerical quantity, without sign.
algebraical condition which this requires,
remark
value
that, as the
is
number
dx.e
2
,
we have
only to
up
to the
of positive errors
and
as the
whole number
of
oo
SIMPLE ERRORS OF OBSERVATION.
positive errors
errors is
is
_
and half the whole number of positive
we must
find the value of
x which makes
f'
1 p dx.e.c
or
1
C \/7T J
For
2!).
dw.e w
equal to j.
*
V 7T Jo
this purpose,
we must be prepared with
dw
table of the numerical values of
V 7r J
u2
'
It is not
our business to describe here the process by which the
numerical values are obtained (and which
the integrals of
all
expressible functions)
give the following table, which
in
Kramp's Refractions and
is
is
we
common
shall
in the Encyclopaedia Metro
Table of the Values of 
VtJo
oo
merely
abstracted from tables
politan, Article Theory of Probabilities.
to
dw.
e"
PROBABLE ERROR.
By
of
interpolation
which gives
0*476948
value of
among
find that the value
value of the integral 025,
for the
or the Probable Error,
is
which is the corresponding
And, conversely,
x 0476948.
x, is c
we
these,
23
= Probable
Error x 2096665.
The reader
30.
table
how
value of
nearly
or
all
will
advantageously remark in this
the errors are included within a small
For
it
be remembered that the Inte
will
gral
when
multiplied by
and negative
or
w = 1*65
or
up
to
49
of the
from
error
that value of
w=
error
number of positive
number of errors up to
(the entire
errors) expresses the
Thus
it
appears that from
= 1*65, we
to
w = 30, we
have obtained
whole number of errors of the same
31.
have already obtained
whole number of errors of the same sign
up
w=
and
49999
f the
K
0nno
sign.
Returning now to the results of the investigations
in Articles 26, 27, 28, 29
we may
conveniently exhibit
the relations between the values of the different constants
therein found, by the following table
24
SIMPLE ERRORS OF OBSERVATION.
PROPORTIONS OF THE DIFFERENT CONSTANTS.
UNUSUALLY LARGE ERRORS.
25
we investigate the value of the modulus, first
by means of the Mean Error, secondly by the Error of
Mean Square, we shall probably obtain discordant results.
We cannot assert a priori which of these is the better.
Thus,
if
There
34.
quently that
is
it
the results of a
be found
one case which occurs in practice so
deserves especial notice.
number
of observations,
it
will frequently
that, while the results of the greater
of observations
are
fre
In collecting
number
very accordant, the result of some
one single observation gives a discordance of large mag
There
nitude.
under these circumstances, a strong
is,
temptation to erase the discordant observation, as having
been manifestly affected by some extraordinary cause of
Yet a consideration
error.
of the law of
Frequency of Error,
as exhibited in the last Section (which recognizes the possible existence of large errors), or
a consideration of the
formation of a complex error by the addition of numerous
simple errors, as in Article 14 (which permits a great
num
ber of simple errors bearing the same sign to be aggregated
by addition
complex
occur
of magnitude,
error), will
and
if so,
and thereby
to
shew that such large
they must be retained.
think that where a cause of unfair error
produce a large
may fairly
may perhaps
may exist (as in
errors
We
omission of clamping a zenithdistancecircle), and where
we know by
certain evidence that in
unfair cause has actually
reason to presume that
stance before us.
come
it
some instances that
into play, there
is
sufficient
has come into play in an in
Such an explanation, however, can only
be admitted with the utmost caution.
COMBINATION OF ERRORS.
2G
PART
ERRORS
II.
THE COMBINATION OF FALLIBLE
IN
MEASURES.
5.
Law of Frequency of Error, and values of Mean
Error and Probable Error, of a symbolical or numerical Multiple
This case
35.
portant that
ment.
of one Fallible Measure.
we
Suppose
exceedingly simple
is
make
shall
it
but
it is
so
im
the object of distinct treat
that, in different
measures of a quantity X,
the actual errors x v x2 x3 &c. have been committed.
Then it is evident that our acceptations of the value of the
,
quantity
Y=nX (an
algebraical or numerical multiple of
X), derived from these different measures, are affected by
y =nx
y2 =nx2 y 3 =nxs &c; and that,
is liable to any number of
the magnitudes x, x 4 Bx, or any thing between
is liable to exactly the same number of errors of
the Actual Errors
errors of
them,
the magnitudes nx
between them.
= y,
The number
nx + nSx =y
+ By,
or of
magnitudes
Therefore the expression for the Frequenc}^
of Errors in Article 24
fall
generally speaking, where
becomes
of errors of
between y and y
+ 8y, may
A
or
c%
nX, whose magnitudes
be expected to be
_**
.
C *JlT
this
Sx,
ERROR OF MULTIPLE OF A MEASURE.
which
27
the same as
is
v
071.
nc V7r
From
this
we
at once derive these conclusions: (1)
law of Frequency of Errors
for the errors of
future combinations,
an
original measure.
nX may
(2)
be used as
The modulus
in the formula, nc.
is,
nX is exactly similar to that
an original measure X; and therefore, in
all
nX
for
(3)
of
nX will
therefore
It
Error
x mean error of X; the Error of
may be
the
useful to guard the reader against one
the measure of a simple quantity
errors (whether actual,
quantity
Mean
the
nX= n x error of mean square of X
of nX= n x probable error of X.
misinterpretation of the meaning of nX.
The
for the errors of
of
Probable Error
36.
=n
had been
if it
Referring to the constant
proportions in Articles 26, 27, 29, 31;
Mean Square
The
X cannot in
We do not mean
Y which
equal to n X.
is
mean, or probable) of the
any way be made subservient
to the
determination of the error of another simple quantity Y.
Thus, reverting to our instances in Article
ment
5,
&c, a judg
of the possible error in estimating the length of a
road about 100 yards long will in no degree aid the judg
ment
of the possible error in estimating the length of a
road about 10000 yards long.
The quantity
nX
is
in fact
merely an algebraical multiple or a numerical multiple of
X, introduced into some algebraical formula, and
is
not
that
we
exhibited as a material quantity.
37.
Another caution
must most
to be observed
carefully distinguish
between
is
this
nX the
multiple
COMBINATION OF ERRORS.
28
of
X (on the one hand), and the sum of a series of n inde+ X + &c. + X (on the other
pendent quantities X,
hand)
...
even though the mean error or probable error of
Xv X
&c. is equal to the mean
The value of mean error or
such a sum will be found hereafter
each of the quantities
error or probable error of X.
probable error of
(Article 53).
6.
Law
of Frequency of Error, and values of
Mean
Error and Probable Error, of a quantity formed by
the algebraical
sum
or difference of two independent
Fallible Measures.
38.
Suppose that we have the number
C of measures
of a quantity X, in which the law of frequency of errors
is
we may expect the number
between x and x + h, to be
this (see Article 24), that
of errors
whose magnitudes
r
fall
c~.li
C\/7T
c
being the modulus of these
errors,
and the number
being very large.
And
suppose that we have the number
F of
measures
of a quantity Y, absolutely independent of the measures
of the quantity
of errors
X,
in
which we may expect the number
fall between y and y + h, to be
whose magnitudes
t
e
being the modulus of these
being very large.
r.h,
errors,
and the number
ERROR OF SUM OF TWO MEASURES.
And
29
Z is formed by adding
suppose that a new quantity
X and Y together.
It
X+
^or
of
required to ascertain the law of frequency of errors
is
Instead of supposing the errors to be graduated,
39.
we
suppose that
will
Y.
all
& h have the uniform magnitude x
.
other equal intervals h in the value of x.
x n x x,
,
number
the errors whose
for
x", for
C7
x 2h,
x h, x + h, x+
the numbers of errors of
C\Jtt
of
magnitude x
of
magnitude x
h,
2h,
and
is
so for
Thus, putting
and putting
2h,
X will be
<?
C'.e a
h;
/i
a" 2
of
magnitude
of
magnitude x +
of
magnitude x
C'.e~c .h;
3
x,
h,
C'.e e.h;
x"
and
it is
plain that, by
+ 2h,
making h small enough,
of things will approach as near as
graduated
Ce'^.h;
we
this state
please to that of
errors.
In like manner, the numbers of errors of
of
magnitude y+2h,
of
magnitude y
+ h,
F'.e
Y will
.h;
 vF'.e f .h;
2
be
COMBINATION OF ERRORS.
30
magnitude
of
magnitude y
h,
F'.e
of
magnitude y
2h,
F' e~
+ y = z.
Now
Let x
40.
values
possible
value of
(F
of
(0
of Z,
F'.e f.h;
y,
v"
.
We
of
all
number
CF
the
possible
of
of combina
(This process implies that the errors of A' are ab
combining the two
result of
form
number) with every possible value
in
solutely independent of those of Y.)
will
h.
in order to
The number
binations of errors will be the same.
Article,
7
we must combine every
in number), forming a total
tions.
f.h;
we
find that the
series of
of com
we examine
If
the
errors in the last
magnitudes of the errors formed
be
4/<,
zSh,
z
shall therefore
Z in
z
2h,
h,
have a
z,z
+ h,z + 2h,
series of
3h, z
ih.
magnitudes of error
our result, varying by a step of magnitude h every
time, and therefore similar to those which
for the errors of
41.
Now
if
X and
we have adopted
Y.
we examine the combinations
that will produce
z,
and the numbers
tions (which apply to a step of
of errors
of these combina
magnitude
h),
we
find the
following:
combining x
2h with y
2h, the
C'.e~*.hxF'.e S'.h
or
number
is
C'F'.hxe'* P.h;
ERROR OF SUM OF TWO MEASURES.
combining x
li
with y
the
4 h,
number
31
is
G'F'.hxtT + 'P.h;
combining x with
number
the
y,
is
C'F' h x e~*~f .h;
.
combining x
+h
h,
with y
the
number
is
C'F'.hx e~^~%.h;
combining x + 2h with y
2h, the
number
is
r
C'F' .h x e~~^~f .h;
and
z
for y,
If
we put
and remark that
y" = y + %h
and
both ways.
so on, continued indefinitely
= z x + 2h z x u
we
so for the others,
see that all the last factors in
the series just exhibited are the values of
when
for
XV
?/2
e~ c2
~f
x we put
x
continued
value of
,32
h,
(ZX)
e~&~ "T*
or
h,
successively the values
h,
2h,
x,
x+
h,
x+
2h,
indefinitely both ways, without
z.
nitely small,
The sum
is
of these, supposing h
altering
made
the
indefi
the same as
+ oo
X?
(ZX)%
 00
where
z is considered constant.
Introducing the
and remarking that
C'F'
F = OF
C
~x/
^,
yvT
7
c\>7r
77"
c 7r
factors,
COMBINATION OF ERRORS.
32
number of errors of magnitude z when a step of
is made each time, or, as in Article 40, the
whole number of errors of Z whose magnitudes are included
the whole
magnitude h
between
and
+ h,
CFh
r+
is
to
_*2 _^r*l2
.
&
/a
be regarded as constant.
The index
42.
7
dx
'J
cfrr
where
be
will
of the exponential
is
easily
changed
into this form;
Let
+/ = /,
2
Then the index
(V +/V
ef
+f
~^= aq~^=
c
z*
is
:,
9'
dx = d%, and z is constant for this investigation), the whole number of errors of Z, whose magnitudes
are included between z and z + h, will be
And,
(as
CFh
cpr
But (see
+ 00

 X
Article 17,
_5
+c
Jy
?
J _oc
and remark that
cf
in this case
qf
ERROR OF SUM OF TWO MEASURES.
number of errors of Z whose
Therefore, finally, the whole
magnitudes are included between z and z
CF
33
_*?
+ h,
will
be
where the whole number of combinations which can form
errors is
CF.
Comparing
43.
this expression
24, it appears that the'
precisely the
with that in Article
law of frequency of error
same as that
for
or for
for
Z is
Y; the modulus
being g or
V(c
Hence we have
two
cally to
).
remarkable
X and Y are
result.
When
added algebrai
form a result Z, the law of frequency of error
Z will
for
+/
this very
determinations
fallible
be the same as for
X or
Y, but the modulus
be formed by the theorem,
will
square of modulus for
Z= square of modulus for X+ square
of
And
44.
modulus
for Y.
Mean
as (see Articles 26, 27, 28, 29, 31) the
Mean
Square, and the Probable Error,
are in all cases expressed
by constant numerical multiples
Error, the Error of
of the
Modulus, we have
(m.e.
(e.
m.
s.
ofZ) 2 =
(m.e.
ofX) 2 +
(m.
e.
of
Y) 2
Zf = (e. m. s. of Xf + (e. m. of Y)\
of Zf =
(p. e. of X) + (p. e. of Y)\
of
s.
(p. e.
A.
COMBINATION OF ERRORS.
34
These are the fundamental theorems
for
the Error of
They
the Result of the Addition of Fallible Measures.
but one theorem
constitute, in fact,
inasmuch
one, the others follow as matter of course.
commonly make use
of Probable Errors (as
sively adopted), unless
the reader,
"but
who
any difference
prefers
Mean
as,
using
"We
shall
most exten
expressly noted
is
Errors,
may form
the
theorems in the corresponding shape, by merely substituting "m.
e." for
"p.e." throughout.
be too strongly enforced on the student
It cannot
45.
that the measures which determine
X must be absolutely
and entirely independent of those which determine Y.
any one of the observations, which contributes
measure of X, does
also contribute to give a
then the single measure of
to
form
its
and the
X with any possi
which the whole investigation in Articles
40 and 41 depends,
tion
measure
single
value of Z, and with no other
freedom to combine any possible error of
ble error of Y, on
measure of Y;
X founded on that observation
must be combined with the corresponding
of
If
to give a
is
to that extent lost.
As an
illustra
suppose that differences of astronomical latitude upon
the earth, or
of the
same
'
amplitudes,' are determined
stars at the
by observations
two extremities of a meridian arc
and suppose that X, the amplitude from a station in the
Isle of
Wight
to a station in Yorkshire, is
observing stars in the Isle of
determined by
Wight and the same
stars in
Yorkshire; and suppose that Y, the amplitude from the
Yorkshire station to a Shetland station,
is
determined by
ERROR OF SUM OF TWO MEASURES.
35
observing stars in Yorkshire and the same stars in Shetland.
First suppose that the observations of stars used
in the
measure of
X are
Then the
the measure of Y.
of
not the same which are used in
errors in the determination
are totally independent of the errors in the deter
bined with any one determination of
Xmay be comY; and if Z= X
+ Y= amplitude
to Shetland,
mination of
Y; any one determination
from
of
Wight
Isle of
the
theorem
(p. e.
of Z)'
(p. e.
Xf + (p.
set of starobservations
made
Y (by comparison
and
of Y)*
in Yorkshire are used to
X (by comparison with
determine
e.
But suppose now that one and the same
applies strictly.
tions)
of
Then the determination
Isle of
Wight observa
with Shetland observations).
of X, based
upon a
starobserva
combined only with a deter
tion in Yorkshire, will be
mination of Y, based upon the same starobservation in
be seen on taking the means of zenith
Yorkshire
(as will
distances
at the stations,
and forming the amplitudes).
The Yorkshire
observations are of no use at
mining Z, and
may be
have no influence on the result
any
star in
tions
make
is
for if
deter
Their errors
the observations of
X too small, the same observaY equally too large, and in forming Z=X+ Y
Yorkshire
make
these errors disappear.
here
all for
completely omitted.
totally
In
fact,
the determination of
independent of those of
investigation of its
depend on those of
mean
X and
Y; and the
error or probable error will not
X and
Y.
It will
depend on the ob
servations at the Isle of Wight and Shetland only: whereas
C2
COMBINATION OF ERRORS.
3G
the probable error of
Isle of
of
X will depend on observations at the
Wight and Yorkshire
will
Shetland
Z= X+
depend on the observations at Yorkshire and
Thus
only.
may happen
it
Y, the probable error of
although
that,
Z is less
than either the
X or the probable error of Y.
probable error of
The
and the probable error
only,
investigation of the probable error of Z,
common
portion of the stars observed are
stations, will
to
be explained hereafter (Article 80).
Suppose that we have determinations of
4G.
Y, as in Article 38, and
W=X Y;
or probable error of
required to
error
W.
The fundamental
upon which we have
supposition,
gone throughout the investigation,
the same
is
is
it
X and
and the mean
ascertain the law of frequency of errors
frequency
when a
two or three
that the
is,
for positive
and
And
is
law of
for negative errors
implied in our
final
formula for the number of errors between x and x
+ Bx,
of the
same magnitude.
namely,
Aj
and
for
positive
jp
c2
= s.
Bx,
which gives equal values
Inasmuch therefore
and negative
equal numbers,
errors as
+ Y;
this
it
follows that
Now
(p. e.
of
r
is
W= X+ (
Wf = (p.
e.
error of
Y.
Y), and therefore
of A')
=+s
liable to
liable to the
and therefore the probable
the same as the probable error of
47.
is
same magnitude
errors of the
as
for
(p. e.
of
Y)\
in
same
7
I " is
ERROR OF DIFFERENCE OF TWO MEASURES.
37
Substituting in the last term from Article 4G,
of
(p. e.
Wf =
The theorems
48.
of
(p. e.
of Article 44
tended, in the following form
[m.
e.
of
(X Y)Y =
m.
s.
of
(X
{e.
{p. e.
and. the
Y)}'
(p. e.
may
F)\
of
therefore be ex
(m.
e.
= (e. m.
(X Y)Y =
of
X) 2 +
s.
(p. e.
law of frequency of errors
of
X) +
of
X) 2 +
of
X) +
e.
of
Y)\
s.
of
F) s
(p. e.
of
Y)\
(m.
(e.
m.
for
X+ Y will be similar
to that for a simple fallible measure.
49.
The
reader's attention is particularly invited to
the following remark.
when the
We
have found in Article 35 that
errors of a fallible
measure are subject to our
general law of Frequency of Errors, the errors of any constant multiple of that measure are subject to the
laws
and we have found in Articles
the errors of each of two
fallible
that law, the errors of their
ject to the
same
law.
44, 47, 48, that
measures are subject to
sums and
Now
all
same
when
differences are sub
our subsequent combina
tions of fallible quantities will consist of sums, differences,
and multiples.
of
which we
Consequently, for every
Frequency of Errors
ficient notice,
we
fallible
treat hereafter, the General
shall
will apply.
Regarding
Law
of
this as suf
shall not again allude to the
Frequency of Errors.
quantity
Law
of
COMBINATION OF ERRORS.
38
Values of
7.
Mean Error and Probable
Error, in
combinations which occur most frequently.
To
50.
find the Probable Error of
kX+ IY,
k and
being constant multipliers.
By
Article 35, the probable error of
and the probable error of IY=1 x probable
Now, considering kX and I Y as two indepen
error of AT;
error of Y.
dent
kX=k x probable
fallible quantities,
{p. e.
of
(kX + 1 Y)}* =
of
(p. e.
kX) 2 +
(p. e.
of
Y)\
Substituting the values just found,
{p.e. of
(7cX+lY)Y = k\(p.e.
of
X) 2 + 1 2
of
X) 2 +
(p. e.
of
Y)\
In like manner,
{m.
of
e.
51.
number
(kX+ I Y)f = k2
To
(m.
e.
(m.
find the Probable Error of the
e.
sum
of
of
F)
2
.
any
of independent fallible results,
B + S+T+ U+&c.
This
is
by repeated applications of the
easily obtained
theorem of Article 44, thus
[p. e.
of
(R +
[p.e.of{(i2
S)}
(p. e.
+ ^) + r}]
{p. e. of (It
= (p. e.
of
of X)
(p. e.
of )
2
;
+ S)Y +
R) 2 +
(p. e.
(p. e.
of
of
S) 2 +
T) %
(p. e.
of
T) s
ERROR OF AGGREGATE OF MEASURES.
[ip.e.o{{(R +
{p. e.
= (p.
and
so
e.
3,9
S+T)+U}Y
of
of
(R +
S+ Tff +
Rf+ (p. e.
of )
(p. e.
Uf
of
+ (p. e.
of T) +(p.
of
e.
U) 2
on to any number of terms.
A similar theorem applies to the Error of Mean Square,
and the Mean Error, substituting
e.
m.
or m.
s.
e.
for p.
e.
throughout.
In like manner, using the theorem of Article
52.
the probable error of
t,
u,
rR + s8+ tT + uU+&c, where
are constant multipliers,
{p. e.
of (rR
is
+ sS + tT + u U)
Mean
Error, substituting
e.
Measures thus combined
s.
and m.
may be
U)\
Mean Square and
a similar theorem for Error of
m.
s,
given by the formula,
=r2 .(p.e. of R)*+s\(p.e. of S) z +t\ (p.e.of T) +w2.(p.e.
And
50,
r,
e.
for p. e.
called "
Cumulative
Measures."
53.
where
To
find the Probable Error of
X Xv Xv ...X
lt
nf
are
X +X
different
. . .
+ Xn
and independent
measures of the same physical quantity, or of equal physical quantities, in
is
every one of which the probable error
the same, and equal to the probable error of
By
the theorem of Article 51,
X
COMBINATION OF ERRORS.
40
(T + X+... + Z)} 2 =(p.e.
{p.e.of
+
= (p. e.
=n
and
of
(p.e.ofX)
of A^)
(p. e.
of
A^+fc.e.
of A',)
...
+ (p. e. of XJ +
2
. . .
+ (p. e. of A^)
n terms
to
X Y;
X
therefore,
p. e. of
(X,
+ X8 + ... + XJ = *Jn
p. e. of
Xv
In Article 35, we found that
54.
nX = n x p. e.
p. e. of
of
x ;
but here we find that
p. e.
of (X,
4
+ X ) = 01
...
of
p. e.
Xv
although the probable error of each of the quantities
&c.
is
equal to that of
Xv A
little
When we
explain this apparent discordance.
the identical quantities
is
consideration will
add together
X Xv Xv &c. to n terms; if there
1}
a large actual error of the
first
Xv there
the same large actual error of each of the other
and the aggregate has the very large actual
necessarily,
is,
X X
&c:
nx
large
error
l5
Xv But when we add together the independent
quantities X X &c,
the actual error of X
large,
error of
t,
is
if
is
it
very improbable that the simultaneous actual error of
each of the others
and the same
sign,
that the aggregate of
A"3 &c, has a value equally large
,
and therefore
all will
it is
or approaching to n x large error of A^.
of the probable error (which
error, see Article 31)
very improbable
produce an actual error equal
is
The magnitude
mean
proportional to the
depends on the probability or
fre
ERROR OF AGGREGATE OF MEASURES.
quency of large actual
mean
error large,
be smaller than that of
of
To
55.
Xj =
X +X +
X +X +
error of
= p. e.
to
. ...
Xv X
where
+ Xn
..
will
terms, although.
of
...
the
large actual errors)
n.
find the probable error of the
X=
p. e. of
make
errors, (for in Article 26, to
we must have many
and therefore the probable
p. e.
41
are
mean
of
different
v
and
independent measures of the same physical quantity, in
every one of which the probable error
The mean
of
and the square of
J (p.
i
(P
J( P
and
XJ +
e f
XiY +
'
its
of
e.
X, I, ...
of
X + X2+~ + X
= p. e.
probable error, by Article 52,
of
XY+
e of
XiY +
(p. e.
"2 (P
. . .
+ i (p. e. of X )\
n
to
w terms,
.e.ofX ) 2 = ^(p.e.ofX r;
1
therefore,
p. e. of
mean
of
Xv X
. .
p. e. of
Xr
SRStTY
'
COMBINATION OF ERRORS.
42
Instances of the Application of these Theorems.
8.
Instance
56.
station
of a
distance
The
(1).
colatitude of a geographical
determined by observing,
is
star
at
on times, the zenith
upper culmination
its
same
observing, n times, the zenithdistance of the
its
lower culmination
being applied.
observations
p. e.
To
1.
is
star at
proper astronomical corrections
all
and by
The probable error of each of the upper
p. e. u. and that of each of the lower is
find the probable error of the determination
of colatitude.
The probable
is
error of the upper zenithdistance,
derived from the
mean
which
of to observations, is
ywi
and the probable
is
error of the lower zenithdistance,
mean
derived from the
Now
of
n observations,
the colatitude
nithdistance
= 9 upper
which
e
tj
is
'
1*
*
zenithdistance
+^
lower ze
and the determinations of these zenith
distances, as facts of observation, are strictly independent.
Therefore,
(p. e.
by Article
of colatitude)
=  (p.
_l
e.
1

'
of u. zen. dist.) 2
(p.e.Ti,)'
52,
to
(p.e.l.)
4
'
(P
e f

zen
dist.)
INSTANCES OF AGGREGATES OF MEASURES.
If the observations at
43
upper and lower culmination are
equally good, so that
p. e. u.
then
(p. e.
or
e. 1.
of colatitude)
p. e.
Instance
57.
= p.
of colatitude
(2).
= p. e.,
= ^'
''
.
= ^
( +
)
mn
In the operation of determining
geographical longitude by transits of the moon, the moon's
rightascension
moon with
is
the
determined by comparing a transit of the
mean
of the transits of several stars
to
find the probable error of the rightascension thus deter
mined.
and
if
If p.
e.
m. be the probable error of moonobservation,
p. e.
s.
the probable error of a starobservation, and
the
number
of starobservations be n, then
p. e. of
p. e.
mean
of star transits
of moontransit
= p.
e.
"'
,*
m.
Hence, by Article 48,
p. e. of
(moontransit
mean
of startransits)
we have
'
44
COMBINATION OF ERRORS.
If p.
e.
s.
= p. e. m. = p. e.,
p. e. of (moontransit
mean
of startransits)
=pVS +1 )It will
amounts
little
be remarked here that, when the number of stars
to three or four, the probable error of result is very
diminished by increasing the number of
stars.
Methods of determining Mean Error and Probable
Error in a given series of observations.
9.
In Articles
58.
26, 27, 28,
we have given methods
Mean Error, Error of Mean Square, and
when the value of every Actual Error in
of determining the
Probable Error,
a series of measures or observations
But
it
is
known.
certainly
evident that this can rarely or perhaps never
is
apply in practice, because the real value of the quantity
measured
is
not certainly known, and therefore the value
of each Actual Error
now undertake
series of
and
is
to find (from the
Error of
not certainly known.
n measures of a physical element
sures being, so far as
good)
is
the solution of this problem.
known
(all
We
shall
Given a
the mea
to the observer, equally
measures only) the Mean Error,
Mean Square, and Probable
mean of the n measures.
Error, of one measure,
of the
59.
sult to
We
shall suppose that (in conformity with a re
be found hereafter, Article GS,) the
mean
of the
CORRECTED DETERMINATION OF MEAN ERROR.
n measures
is
adopted as the true
is
not necessarily the true result
result.
;
Yet
4.)
mean
this
and our investigation
will naturally take the shape of ascertaining
how much
the formulEe of Articles 26, 27, 28, are altered by recognizing
its
chance of
And
error.
for
first,
Mean
In
Error.
the process of Article 26, suppose that, in consequence of
our taking an erroneous value for the true result,
+ errors are increased
errors are diminished
Then the mean
error
by a small quantity, and
(numerically)
And
the
errors
if,
is
to
this
is
different
is
+ errors,
not affected
is
sum
if,
found in
Mean
Suppose
Square.
a, b, c, d,
&c.
is
+ h + c + d + &c.
~
n
and therefore
of
2G
A result may follow from
then the Actual Error of the mean
the
of Article
been remarked in Article 33.
Secondly, for Error of
n terms
error
Thus the determination
that the Actual Errors of the n measures are
to
not be
mean +
slightly inconsistent with that to be
Article 60, as has
60.
the
and the process
be used without alteration.
which
error, will
ways (numerically), and their
sensibly unaffected.
Error
error will be, one
error are very nearly equally affected in
magnitude but in
mean
Mean
the
from the same cause, one or more of
become apparently
and the mean
all
by the same quantity,
and their mean, which forms the mean
affected.
the
by the same quantity.
and the mean
increased and the other diminished,
all
for
the process of Article 27,
of the squares of the
we form
Apparent Error of each mea
COMBINATION OF ERRORS.
46
each measure from the
sure, that is of the difference of
mean
we do not form the squares
of a,
b, c, d,
&c, but of
a+b+c+d+ &c.
,
a+b+c+d+ &c.
+ b + c + d Jr Sec
a
c
The sum
of their squares (that
cf apparent errors)
is,
sum
the
of
is
+l +
c"
+ &c
~(a + b + c + &c.) x(a + b + c + d + &c.)
+ nx x(a + b + c + d + &c.Y
= a + b + c + &c 2
Now,
in the long
(a
K
+ b + c + d + &e.
run of observations,
each of the squares in the
first
or o
or c
Mean Square of Error
&c, we may put (Error of Mean
the definition of Article 27.
may
consider
part of this formula as
being equal to the
2
we
'.
'
But
for
so that for
a",
Square)* using
+b+c+d
which enters as an aggregate quantity, we must remark
that,
by Article
51,
CORRECTED DETERMINATION OF PROBABLE ERROR.
Mean Square
= (m.
n x
(e.
v
m.
(Error of
of a measure)
s.
And from
,
m.
s.
e.
m.
s.
of a
ot
And by
(e.
s. e.
m.
Square)
we form
xn x
s.
of b)
(e.
+
.
truly
is
m.
&c.
of a measure)
s.
of a measure)
2
,
2
.
this,
measure
the
+ b + c + d + &c.)
Mean
'
e.
+ (m.
of squares which
= (n 1)
p. e.
of a)
s. e.
=n x
Thus the sum
of Error of (a
47
mean =
/sum
of squares of apparent errors
^
711
/sum

V/
.
of squares of apparent errors
7i
.
(n
1)
the table of Article 31,
of a measure
= 06745
p. e. of
the
'sum of squares of apparent errors
A
mean
O'o/4o
sum
4
,
of squares of apparent errors
A /
7i
[n
=r
1)
The quantities which enter into the formation of
mean error, error of mean square, and probable error,
61.
the
will
be most conveniently computed thus.
that the different measures are A, B,
their
mean
is
M.
C,
It
is
supposed
&c, and that
COMBINATION OF ERRORS.
48
the
First, for
mean
Select
error.
M:
A, B, C, &c. which are larger than
number
to be
I,
the measures
all
supposing their
form the quantity
A + B+G+&C.
Mr
,
'
which gives one value of mean
error.
Select
measures P, Q, B, &c, which are smaller than
posing their
number
to be
all
M\
the
sup
form the quantity
s,
M P+Q + B + &C.
s
which gives the other value of mean
mean
these two values of
error
Second, for the error of
is
mean
to
error.
The mean
of
be adopted.
square and probable error.
 Mf +(B Mf +(C M) + &c.
This = A + +6' + &c.2J/. (A + B+C + &c.)+n.3r.
We
wish to form (A
2
A + B+C + &c. = n.M;
But
so that the expression
= A +B*+
2
This
is
the
"Sum
C"
+ &c.  n M\
.
of squares of apparent errors," to be
used in the formula) of Article 60.
USE OF COMBINATIONWEIGHTS.
PART
49
III.
PRINCIPLES OF FORMING THE MOST ADVANTAGEOUS
COMBINATION OF FALLIBLE MEASURES.
Method of combining measures; meaning of "com
10.
binationweight ;"
combination
principle
caution in
its
of most
advantageous
application to " entangled
measures."
62.
The
determinations of physical elements from
numerous observations,
to
which
this treatise relates, are
of two kinds.
The
First
is,
the determination of some one physical
element, which does not vary or which varies only by
a certainly calculable quantity during the
observations,
measures.
by means
of
numerous
direct
period
Thus, in the measure of the apparent angular
distance between the components of a double star,
making
of
and immediate
direct
we
are
and immediate measures of a quantity
sensibly invariable; in measuring the difference of moon's
right ascension from the right ascension of
known
two or more known
render similar
observations at an
ing
its
longitude,
stations, in order to
unknown station
we are making
stars at
available for determindirect
and immediate
measures of quantities which are different at the two or
more
stations,
but whose difference can be accurately com
puted.
A.
ADVANTAGEOUS COMBINATION OF MEASURES.
50
The measures thus obtained
63.
the problem before us
is
is,
How
are all fallible, and
they shall be combined
It
not inconceivable that different rules might be adopted
for this purpose,
depending
(for instance)
upon the products
of different powers of the various measures, and ultimate
But the one method
in effect)
sum
of the root corresponding to the
extraction
indices of powers: or
upon other imaginable
(to
which
all
others will approximate
which has universally recommended
only by
its simplicity,
permits
all
of the
operations.
not
itself,
but also by the circumstance that
it
the measures to be increased or diminished by
the same quantity (which
is
sometimes convenient),
is,
to
number (either different for eacli
the same for any or all) which number
multiply each measure by a
different measure, or
is
here called the "combinationweight;" to add together
these products of measures by combinationweights; and
sum by
to divide the
the
The problem
64.
becomes
combinationweights
Arid to answer
the criterion of advantage.
shall
fix
result
is:
That
correct
is
this,
The
combination
whose probable
mean square,
we can do.
of combinationweights.
of advantageous combination
What
this,
advantageous
sum
error,
or
we must
criterion
is
decide on
on which we
best which gives a
mean
error, or error of
the smallest possible.
We
now
be most
will
This
is
all
that
cannot assert that our result shall be
or that, in the case before us, its actual error shall
be small, or smaller than might be given by
combinations; but Ave can assert that
it is
actual error will be the smallest, and that
many
other
probable that
it is
its
certain that,
PROBABLE ERROR OF RESULT TO BE MINIMUM.
by adopting
we
this rule in a
very great number of instances,
on the whole obtain results which are
shall
51
liable to
smaller errors than can be obtained in any other way.
Now
65.
if
we know
the probable errors, or the pro
portion of probable errors, of the individual observations,
(an indispensable condition,)
for
we can put known symbols
them, and we can put undetermined symbols
for the
combinationweights; and, by the precepts of Part
II,
we
can form the symbolical expression for the probable error
of the result.
This probable error
mum, the undetermined
is
Thus we fall upon the theory
weights.
and minima.
to
be made mini
quantities being the combination
Its application
is
of complex
maxima
in every case very easy,
because the quantities required enter only to the second
order.
Instances will be found in Articles 68 to 72.
66.
It
sometimes happens
that,
even in the measures
an invariable quantity, combinations of a complicated
Different complex measures are somecharacter occur.
of
times formed, leading to the same result; in which some
of the observations are different in each measure, but
other observations are used in
all
or in several of the
measures; and thus the measures are not strictly independent.
We
shall
The only caution
call
these
"entangled
to be impressed
measures."
on the reader
is,
to
be
very careful, in forming the separate results, to delay the
exhibition of their probable errors to the last possible
stage; expressing first the actual error of each separate
result of the
form ultimately required, by the actual error
D2
ADVANTAGEOUS COMBINATION OF MEASURES.
52
of each observation.
It will often
be found
that, in this
way, the results of observations will be totally or partially
eliminated (and justly
had been formed
the result.
(Articles
74
at
Instances
would have vitiated
given
below
the simultaneous
deter
this
of
the probable errors
if
stage,
be
will
to 85).
The Second
G7.
which,
so),
an early
class
is,
mination of several physical elements.
trated by one of
its
It
may be
illus
most frequent applications, that of
determining the corrections to be applied to the orbital
elements of a planet's
right ascensions
the planet
is
orbit.
The
and north polar
at
different
quantities measured are
when
distances, observed
points
in
its
and
orbit,
different positions with respect to the observer.
in
If ap
proximate orbital elements are adopted, each having an
indeterminate symbol attached to
which
it
may
require;
it
will
it for
the small correction
be possible to express, by
orbital calculation, every right ascension
distance
by numerical
quantities, to
and north polar
which are attached
definite multiples of the several indeterminate symbols.
Equating these to the observed right ascensions and ninth
polar distances, a long series of numerous equations
is
obtained, with different multiples of the indeterminate
symbols; each equation being subject to
error of observation.
How
shall these
form exactly as
And
its
own
the question before us
numerous equations be combined
many
equations as the
number
actual
is
now,
so as to
of indeter
minate symbols, securing at the same time the condition
that the probable error of every one of the values thus ob
COMBINATION OF SIMPLE MEASURES.
tained shall be the smallest possible?
maxima and minima.
of complex
This
is
53
also a case
Numerous problems
in
astronomy, geodesy, and other applied sciences, require this
treatment.
be fully explained in Articles 87 to 122.
It will
Combination of simple measures; meaning of "theoretical weight;" simplicity of results for theoretical
11.
weight; allowable departure from the strict rules.
tween two
a priori;
Let
we have n independent measures
Supj)ose that
68.
some element
of observation
[e.g.
equally good, so far as
stars), all
method
to find the proper
Ev E ,...EM be the
of
the angular distance be
of
we can judge
combining them.
actual errors of the individual
measures, which are not known, but which will affect the
Let their probable errors be
result.
which = e.
wv
as
w.
is
...
And
wn
Then the
ev
e2
...
en ,
each of
the combinationweights required be
let
actual error of the result, formed
described in Article 63, will be
w E + w E ... + wn En
w + w + ... +wn
1
w +w
1
The
(p. e.
...+
+ wn
...
of result)
w + wz
i
2
,
'
nJ
this instance
\
w +w
1
by Article
(w t
which in
\wt
+w
...+
52, is
+w
. . .
...
+ wny
becomes
l
tu n
+W
&2+&C
ADVANTAGEOUS COMBINATION OF MEASURES.
54
Making the
minimum with respect to w v we obtain
fraction
2w,
W* + W* ... + W n
Similarly,
W^ + M?2
2w
w + iu
+ w*
...
It follows that
That
m/j
is,
+ wn
= 0,
=w =w
2
metical
&c, but that
all
are inde
the measures are to be combined by
equal combinationweights;
in other words, the arith
or,
mean is to be taken. The
(probable error of result) 2
=V
or'
probable error of result
as
so for the other weights.
terminate.
= 0.
by w#
w* + w 2
and
+ wn
. . .
was found in Article
69.
results
= r
55.
Suppose that we have n independent measures or
which are not equally good.
(For instance: the
atmospheric or other circumstances of individual observations
may be
different
individual observations are
or, if
equally good, the results of different days, formed by the
means
of different
ferent days,
tions of colatitude
of results
numbers of observations on the
would have
different values.
by means
from different
The
of different stars, the values
stars will
polar distances, as well as
(w x
may be
retained, rejecting
Thus we have for
e.
w*e* +
be affected by their north
by the other circumstances.)
notations of Article G8
only the simple letter
vr*e* ...+
dif
In determina
w.,
f
wn
wny
(p. e.
of result)
8
,
THEORETICAL WEIGHT.
wv w
and
55
&c, are to be so determined as to make
this
minimum.
Differentiating with respect to
2"A'
*i
+ w2 ea
+ wn
wv
_1
^,
Differentiating with respect to
+ lU ...+Wn
tv
And
+ iv *e
'
... I
wn\"
J^ + w
w +w
l
...
=a
n
so for the others.
It
is
evident that w^e*
= w e* = &c. = w n e^ = G
2
CO
(p.e. of result)
_ C(w
70.
...+
We
of result)'
shall
now
w _
tl
(p. e.
+w
Or
some
Hence
indeterminate constant.
and
2w 2 e2*
w{e*
=0
'
C_
111
e*
e*
introduce a
H...H
en
new
term.
Let
(probable
error)*'
be called the "theoretical weight," or
t.
Then we have
w.
these two remarkable results:
When
independent
fallible
measures are
collateral, that
is, when each of them gives a measure of the same unknown quantity, which measures are to be combined by
combinationweights in order to obtain a
final result
ADVANTAGEOUS COMBINATION OF MEASURES.
5G
The
First.
combinationweight, for cacli measure ought
to be proportional to its theoretical weight.
When
Second.
sure
the combination weight for each mea
proportional to
is
its
weight of the final result
theoretical weight, the theoretical
is
equal to the
retical weights of the several collateral
When
sum of the
measures
theo
1
.
the theoretical weights of the original fallible
measures are equal, and they are combined with equal
combinationweights, the theoretical weight of the result
is
proportional to the
number
of the original measures.
These rules apply in every case of combination of
71.
measures leading to the value of the same simple quantity,
provided that the observations on which those measures are
founded are absolutely independent.
Thus,
we may com
bine by these rules the measures of distance or position of
double stars
made on
the same star
the results
(for
(for
different stars
different days; the zenith distances of
geographical latitude) on different days
geographical latitude) of the observations of
the results
(for
geodetic amplitude) of the
observations of different stars ; the results (for terrestrial
longitudes) of transits of the
Instance.
72.
moon on
In Article 56
different days, &c.
we have found
probable error of colatitude determined by
1
The
reader
is
cautioned, while remembering these important theo
rems, also to bear in
When
is,
for the
m observations
mind the
independent
following (Articles 44 to 52)
measures or quantities are cumulative, that
when they are to be combined by addition or subtraction to form a new
fallible
fallible
quantity; then the square of probable error of the new fallible
is equal to the sum of the squares of probable errors of the several
quantity
cumulative measures or quantities.
'
INSTANCE, DETERMINATION OF COLATITUDE.
of a star at
its
57
upper culmination, and n observations at
its
lower culmination,
/m + n
~2\
where
mn
'
probable error of an observation,
e is the
Another
supposed equally good.
are equally good, observed
star,
all being
whose observations
times at upper and n l times
at lower culmination, gives a result with probable error
e
/?,
+ Wj
lujT^
a third gives a result with probable error
a
L A /
&c
>
in sn 2
Their theoretical weights are
4
J
e'
The
'
mn
m+n
4
'
different results
mean)
with
m,n,
m +
'
nt
'
e"
2
*,
m +n
'
ought to be combined
combinationweights
&c.
(to
proportional
form a
respec
tively to
mn
m+n
mna
+ n2
a
m,n,
+n
and the theoretical weight of the mean
and
mn
e~
\m + n
mn
l
vn,
yi a
+n
formed will be
so
m +n
%
&c .;
'
probable error will be the square root of the re
its
ciprocal of this quantity.
It
error.
is
supposed here that the zenithpoint
If
it
is
not, the case
becomes one of
observations," similar to that of Article 75.
is
"
free
from
entangled
+
ADVANTAGEOUS COMBINATION OF MEASUEES.
58
We may
73.
however depart somewhat from the
precise rule of combination laid
down
out materially vitiating our results.
in Article 70, with
We
determined the conditions which make
mum
have in Article
p. e.
Gi)
of result mini
and it is well known that, in all cases of algebraical
minimum, the primary variable may be altered through a
;
considerable range, without giving a value of the derived
much differing from the minimum.
we had two independent measures,
function
pose that
Thus, supfor the
physical element, whose probable errors were e and
We
ought, by the rule of Article 70, to combine
combinationweights in the proportion of 4
pose that
we use combinationweights
Put
1.
and E'
error of result will
n
the p.
e.
will
n
n
for the
1.
same
e'
= 2e.
them by
But sup
in the proportion of
actual errors
the actual
be
n+1
n+1
be (by Article 52)
V e+

l)
V e J e VOr + 4)
\= nr+ r
[n
+ V
32m
/f
Using special numbers, we find
With combinationweights
as 2
1,
the
p. e.
of result
= x
cf
o
0943.
RELAXATION OF RULE.
With combinationweights
as
4:1, the
p. e. of result
59
ADVANTAGEOUS COMBINATION OF MEASURES.
60
Treatment of Entangled Measures.
12.
The nature and treatment
74.
will
of entangled measures
be best understood from instances.
Instance
station
moon
is
to
(1).
Suppose that the longitude of an unknown
be determined by the right ascension of the
at transit (as found
by ascertaining the
tween the moon's time of transit
of transit of
stars)
same manner
(where the number of stars observed
Then, as has been found in Article
right ascension at the
by Articles 47 and
unknown
station
is e
at a
of the times
or of
any
star to
is e
and
+
/
a/CHA
48, as these
at
station
be
e.
probable error of
57, the
station
known
and suppose the
is a);
moon
probable error of transit of the
known
difference be
mean
compared with the right ascension
transit determined in the
that at the
the
arid
/(+
therefore
two determinations are
every respect independent, the probable error of the
in
differ
ence of right ascensions at transit (on which the longitude
depends)
iseW^
Supposc that a second comparison
transits at the
and
unknown
b stars at a second
error of the quantity on
station,
is
made, of the same
with transits of the
known
moon
The probable
station.
which the longitude depends
found in like manner to be
e A
+j+
is
ENTANGLED MEASURES.
Now
if
we combined
these two results, (leading to the
same physical determination, and both
we should
of Article 70,
Gl
correct,)
by the rules
obtain an erroneous conclusion.
OS.
For, the
two
results are not independent,
observations at the
To
75.
unknown
station enter into both.
we must
obtain a correct result,
In
actual errors.
strictness,
inasmuch as the
we ought to
error of each individual observation
refer to the *
refer to the actual
but,
inasmuch as
perfectly certain that all the observations at each
stations, separately considered, are entirely
all
it is
of the
independent of
we may put a symmoon's Ii.A. above mean
the observations atjihe other stations,
bol for the aatiim error of excess of
Let these symbols
of stars' R.A. at each of the stations.
Then the
be N, A, B, respectively.
aefcfc*=eErors
quantities on which longitude depends, as found
paring the
of tin
by com
unknown station with each of the known stations,
are respectively
N A, N B.
Let the quantities be
combined with the combinationweights
Then the
a, /3.
hnal &e&ral error will be
And
+ /3
+ j3
+ /S
the square of probable error of final result
K ^ + dw
e of

To make
this
(p e

0{Ay
(=.
w?
minimum, w e must make
r
a (p.
e.
of A)'
+ /3
(p. e.
of
Bf
(p e of *>'}

ADVANTAGEOUS COMBINATION OF MEASURES.
G2
This algebraical problem
minimum.
as that of Article 69, and the result
exactly the same
is
is
G
R/3
"~(p.e.of)
"fae.ofAy
G is
where
(p. e. of
,
Nf
+
;e
(l
we put
^1
Af x(p.e. of
of
+ ^e.of^)4
2
(p.e. of
(1
If
this gives for
70.
'
{(p.e.ofAT
And
an indeterminate constant.
(p. e. of final result)
S!
q)(l
) )
6)
r for the "theoretical
result (see the definition in Article 70)
of the observations
weight" of
final
n, a, b, for
those
N, A, B, respectively; then the
last
formula but one becomes
11
_
1_
(a
+b
'
+ b) n
n+(afb)*
Let n be divided into two parts n a and n b such that
,
na
Now
if
+b
n,
'
n,b
b
,
n.
atb
the theoretical weight n a at the station
N had
more complicated character may be seen in the Memoirs of the E. Astronomical Society, Vol. xix. p. 213.
* Instances of a
PARTITION OF THEORETICAL WEIGHT.
63
been combined with the theoretical weight a at the station
A, they would have given
weight of their
for theoretical
result
a'n
a
r.
Da
+a
an
a
And
if
+b
n.
an
(a
'
b)
+b
the theoretical weight n b at the station
N had
been combined with the theoretical weight b at the station
B, they would have given for theoretical weight of their
result
b'n
r,
nb + b
_bn
a
And
+b
bn
.
(a
'
b)
consequently,
ra
And
to
+b
n,,
be
it is
+ rb = r.
easy to see that, as there are two conditions
satisfied
by the two quantities n a n b
,
quantities will produce the
77.
weight n
Hence we may conceive that the
is
no other
same aggregates n and
r.
theoretical
divided into two parts proportional to a and b,
and that those parts are combined separately with a and b
respectively,
and that they produce
rate parts ra
theoretical
D, &c.
rb
if
in the result the sepa
which united make up the entire
The same, it would be
any number of stations A, B,
weight of result
found, applies
C,
and
there are
r.
ADVANTAGEOUS COMBINATION OF MEASURES.
G4<
The
78.
partition of theoretical weight of final result
thus obtained, producing separate theoretical weights of
with
parison of
N with
Hence
AT with
A, and com
moon's place) to which
(as
inferred from that to
relates.
and
B, without necessarily distinguishing
whether the element
element to which
does in fact produce separate
respectively,
theoretical weights for comparison of
is
N with
depending on the combination of
result
which
relates, or
relates is inferred
relates
whether the
from that to which
applicable to such cases as the
is
it
JSf
following.
Instance
79.
(2).
geodetic theodolite being con
sidered immoveable, observations (whose actual error
M)
are
dian,
made with
it
for the direction
and observations (subject
&c.) are
made on
is
of the north meri
A, B, C,
to actual errors
different triangulationsignals
to find
the weight to be given to the determination of the true
azimuth
of each signal.
Using analogous notation, the theoretical weight
to
m m m
be divided into parts
a,
b,
c,
&c.
m is
and then the
weights of the determinations for separate signals are
those produced by combining
with
a,
with
b,
&c,
or are
bm
am
m + (a + b + c &c.)
80.
Instance
(3).
'
m + (a + b + c &c.)
&c.
In the observation of zenithdis
tances of stars for the amplitude of a meridian arc divided
f^S.
ENTANGLED MEASURES.
65
two sections by an intermediate station
into
a stars are observed
at all the stations, the
A A A
errors being respectively
are observed at the
means
of the
first
means
of actual
suppose that
and second
errors being
actual
suppose that
b stars
stations only, the
respectively
B B
t
that c stars are observed at the second and third only,
the means of actual errors being
observed at the
are
and third
first
actual errors being J) 1
C C and
only,
d stars
the means of
that
They may be represented
thus
oj
Ti
o3
fc
a>
o3
3
S
" "S
o
h 
<K
id
O _^
w
O g
m
Or^
J**
I
U2
I02
4 3
OQ
r
!h w
4^
^4 2 ,
Stars 6
Bv
Stars c
cZ
c3
aj.jj
Stars a
Stars
a
?o
"1
S
.<
a>
x,
Suppose the probable error of every individual observation to be
e.
It is
now
required to find the combina
tion proper for determining the amplitude of the
first
section of the arc.
81.
Besides the direct measures of the
first section,
there are indirect measures produced by subtracting the
measures of the second section from the measures of the
OG
ADVANTAGEOUS COMBINATION OF MEASURES.
whole
arc.
All the possible measures of the
first
section
are therefore the following
A A
I.
II.
Bt
A^(c ay[
iv.
D  Dl (A A )\
D D {C Cj]
V.
VI.
of these, III
the four measures
a mere reproduction of I
is
IV, V, VI,
I,
may be formed from
all
\ BireCt
4, ^(4, 4,)]
in.
But
BJ
it is
the three others
would introduce indeterminate
may be
B B
and of
one
and the retention of
solutions.
The following
retained, as substantially different
A A
A
easily seen that
with combinationweight
w,
AC +C
3
x,
D D C +C
z
v,
y.
These are entangled measures, inasmuch as
A lt C C
2
appear in different measures.
The
82.
actual error of their
mean
will
v(A t A )+to{Bt B ) + x (A z A,Ca +Ct )+y
v + w+x + y
l
= {v+x)A +vA
1
be
(BD C +
3
< [)
+xAwB +wB +(x+ y) C (x+y) C.yD.+ yD,
v+w+x+y
1
ENTANGLED MEASURES.
The independent
so for the others,
p. e.
of resultN
we
are
fallible quantities
and, by Article 52, remarking that
67
(p. e.
now
of
separated
A f= ,
l
and
find
(y
Making
this
+ w + x+yf
minimum
with regard to
v,
w, x, y, as in
Article 69,
(v
v
=
+ x)+v'a
a
C,
11
W r +W T =
(y
+ x)
'a
(J,
+ x  + {x+y)  + (x +y)
a
C,
=g
{x+y)\+ {*+y)]+y\+y\
From which
2
_ 4a +
2ac + 4<ad
~8a +
+ 6^T
4a& + She + Sbd
b n
W ~2 C
'
 8a + 6c + 6d n
2ac 2nd
X
~8a + Gc + 6d
6c
'
'
2ad + Scd
~ 8a + 6c + 6d
E2
ADVANTAGEOUS COMBINATION OF MEASURES.
68
It
is
remarkable here that in some cases x
negative, indicating that advantage will be gained
may
be
by sub
tracting that multiple of measure from the others.
If a
= b = c = d,
and
D = aC,
the combination weights
become
D
83.
'
we thought
If
fit
'
to reject the combination
A G +G
a>
there would be no entanglement
and
it
would
easily be
found, by Article 70, that the combinationweights ought
to
cd
be proportional to
a, b,
and the
theoretical weight
of the result
fa
cd
In like manner, for the second section of the
arc,
the
measures to be used are
a a
3
a
and the theoretical weight of
1 (a
'
84.
Now
if Ave
zvA'B.+tfi;
result
bd
2b
+ 2d.
combined these two sections
to
form
the whole arc, and inferred the probable error of the whole
from the probable errors of the sections by the rule of
ENTANGLED MEASURES.
Article 44,
we should
69
obtain an erroneous result.
For, the
observations on which the determinations of value of the
two
sections are
founded are not independent both contain
the observations
A B
,
B^, C,
D D
t
and they are
therefore entangled results.
The
correct result for the
whole
an investigation exactly similar
There
A A
3
Dj
the direct measure by the
and the indirect measure obtained by adding
stars,
the result of the b stars to the result of the
B B + C Cr
error
1 fa
is
with error
c stars,
with
theoretical weight of the
be found to be
?
If the
The
result will
tion
part.
the direct measure by the a stars, with error
is
L\
be obtained by
will
to that for each
number
\2
+2+
be
26
+ 2c,
of observations at the intermediate sta
very small,
(as if
is
small, b
and
= 0, d
large,)
the theoretical weight of the value of each section will be
small, while that of the entire arc
This instance
is
may be
great.
well adapted to give the reader a clear
idea of the characteristic difference between actual error
and probable
if
error.
So
we add the measure
to the
we
far as actual error is concerned,
of one section with its actual error,
measure of the other section with
entirely (and correctly) destroy so
its
much
actual error,
of the actual
error as depends on the observations at the intermediate
station.
But the probable
error (see Article 8)
is
a mea
ADVANTAGEOUS COMBINATION OF MEASURES.
70
sure of uncertainty; and
if,
without looking carefully in
each case to the origin of the uncertainty,
we simply add
together the two separate measures charged with their
respective uncertainties,
we
sum with an apparently
obtain for the whole arc a
large imc&rtainty which
is
very
incorrect.
If the observations at the three stations are to be
85.
combined in one connected system
it
will
be best to use
each batch of stars separately, giving to each resulting
amplitude
only.
proper weight as deduced from that batch
For the batches B,
clear;
7G,
its
for
must be
79,
C,
D, the operation
is
perfectly
the batch A, the principles of Articles 75,
used, which here give a very simple
result.
86.
It
is
scarcely necessary to delay longer on the
The caution
subject of entangled measures.
which
in all cases suffices,
is
required,
gations by the use not of probable but of actual errors
collect
all
the
coefficients of
and
to commence the investieach actual
error,
to
and to
them from the coefficients of every other error
and when the formulae are in a state fit for the introducseparate
tion of probable errors, to investigate,
to
by a process
special
the case under consideration, the magnitudes of the
combinationweights which will produce the
probable error in the result.
minimum
DETERMINATION OF SEVERAL ELEMENTS.
Treatment of numerous
13.
unknown
several
"
minimum
87.
The
quantities
(the
to
introduction of the term
origin of equations of this class has been
&c,
x, y,
&c. which
applying
equations
squares."
explained in Article G7.
putting
71
been seen that,
the object of the problem
it is
number
It has there
for the corrections to orbital elements,
of
which elements we
discover,
to
for clearness
shall
suppose to be three, though the investigation will evidently apply in the same form to any
corrections,) every equation will
ax + by +
where
of such
cz
=f
the difference between a quantity computed
is
theoretically
served,
number
have the form
and
observation.
from assumed elements and a quantity obis
therefore subject to
the casual error of
terms of the equations, as given
If the last
immediately by observation, have not the same probable
error,
we
shall suppose that the equations are multiplied
by proper
factors (see Article 35), so that in every case
the probable error of the last term
an arbitrary number,
for
f is
made =
which sometimes
venient to substitute the abstract value
the letters
a, b,
c,f and
we
it is
We
1.
being
very conshall use
others which are to be introduced,
without subscripts, in their general sense
separate equations
but for the
shall affix the subscripts 1, 2, &c.
ADVANTAGEOUS COMBINATION OF MEASURES.
/2
88.
three,
The number
and
to three in
it
of equations being greater than
being requisite to reduce the
number the only method which
;
for giving every
equations
final
suggests
share in the formation of those three equations,
by a
to multiply the equations
and to adopt
their
sum
series of factors
series of
alx
fundamental equations
+ b y + clz=f
l,
aix+bay + c^=fs
&c.
form the three
series
h^x + k&y + \c z = \f
x
h2 a 2x
y ,
+ hp y + k,c,z = k.,f
&c.
l/i.jc+l2 b 2 y+12 c2 z
= lj],
&c.
m^x + vrijbjy + m^z = m^,
max+m b y+ m cz =m f
2
first
h1} k 2 &c,
them by another series l 1}
multiply them by another series m,,
Thus having the
Ave
is
as one fundamental equation
secondly, to multiply
thirdly, to
itself,
one of the fundamental equations a proper
&c.
2 2
2,
&c.
m, &c
DETERMINATION OF SEVERAL ELEMENTS.
of which the
sums are
x t
(Jca)
x X
(la)
+ y 2 (kh) + z.t (Ice) = 2 (A/),
+ y.S (lb) +z.Z (Ic) = X (If),
.
2 (ma) + y S (mb)+z 2 (mc) = S
.
These are our three
x, y,
and
final
z,
for
h,
(mf).
equations for determining
and our problem now
of formation of the factors
of x, y,
73
I,
to ascertain the
is,
law
m, which will give values
may
each of which the probable error
be
minimum.
89.
Let us confine our attention,
for a short
to the investigation of the value of x.
The
time,
process of
solving the last three equations will consist, in fact, in
finding different factors wherewith the equations
multiplied, such that,
when
may be
the multiplied equations are
added together, y and z may be eliminated, and the terms
depending on x and /may alone remain. But, remarking
how the
three equations are composed from the original
equations, this multiplication of equations formed
of multiples of the original equations
of
sums
is
by sums
in fact a collection
of other multiples 'of the original equations.
Let
n be the general letter for the multipliers (formed by
this
final
double process) of the original equations
process for
solution
of the equations
hibited;
xX (na) = 2 (nf)
t{nb)=0;
$(nc)=0;
then the
is
thus ex
ADVANTAGEOUS COMBINATION OF MEASURES.
74
which can be solved with an
infinity of different values
of n.
From
90.
these,
= tt / +j ,+&c..
a + n a. + &c.
p
S(n/)
'
(na)
??
z
2
from which the actual error of x
1a 1
x actual error of f
+ >i a + &c
2
a +
1
n 2a 2
w x actual error
+ &c.
ofyfa2
+ &c,
and, as the probable error of each of the quantities/^,
&c.
= e,
the square of probable error of
ni
e~
(7*
= e"
The numbers
square
nv
its
+ &c
&c. are so to be chosen that the
them
we put Bnt 8n 2
(if)
shall
variation produced
variations in each of
of probable error of
therefore
If
v.
"2
a +?? 2 a 2 + &c.)
1
shall
be
be
minimum
0.
&c. for such small variations,
must have, by the formulae
and
by simultaneous small
of ordinary differentiation,
we
~0
DETERMINATION OF SEVERAL ELEMENTS.
_ n^Sw, + nj>n + &c. _ a
2
+ n* + &c.
Wj"
n,Sn, +n
~
^
0=
l
or
8n a2
Sn
w^j
+ &c.
a,$n,
+ a Sn + &c.
+ n a + &c.
2
a8i>
'
&c.
....
rin
L1J.
here, as
variations 6^, S 2 &c, are not independent
were the corresponding variations in Articles 68
and 69
for
But the
they are affected by the antecedent conditions
= 0, 2
(nb)
(mc)
=
=
from which we derive
b 8n 1
l
b 2 Sn 2
+ &c
= 0^ + c 8n +
2
[2],
&c
[3].
These three equations must hold simultaneously
the values of nv n 2 &c., which
,
91.
It
We
Let h = av
the conditions.
m =b m =b
1
to ex
from these equations the values of n v n 2
are however able to
certain form given to the
require.
would perhaps be a troublesome matter
tract analytically
&c.
we
for
shew
numbers n v n 2
k2
=a
2>
synthetically that a
&c.
&c. satisfies
=b
b &c;
2
&c.; so that the final equations of Article
88 take the form
x t
+ y.S
+ z.X
=2
(a/)
[4],
+ y.Z(b"~)+z.Z(bc)=S(bf)
[>],
x.Z(ac)+y.Z(bc)+z.Z(c*) =S(c/)
[6].
(a
x.S(ab)
Then the
(ab)
values of
x, y, z,
(ac)
which are deduced from
these equations, possess the properties required.
ADVANTAGEOUS COMBINATION OF MEASURES.
7f>
multiplying the
by
third
being
r,
[p
first
obtain the value of x by
by
of these
+q 2
.
coefficients of
q,
the
y and
.3
Then
to vanish.
(a
the second by
p,
and taking their sum, the
made
x x
we
For, suppose that
92.
+r 2
(a b)
(ac)}
p.S(a/) + 2 .S(J/) + r.X(o/};
p.S.(a6)
l>
+ 2.X(6s) + r.S(&c) = 0;
+ ^. 2 (Jc) + r.2
(c)
(c
=0;
which are the same as
pa+ qb + re)} = 2 [f(pa + qb+ re)} .... [7],
${b(pa+qb+rc)}=0
[8],
xx 2
(a
2{c(pa + qb + rc)} =
[9].
Comparing these equations with those
is
now
+ re.
replaced by
pa +
qb
=2
{(pa
+ qb + re)
(n~)
= p 2 (an) + q 2 [b [pa + qb +
.
The
last
of Article 89, n
Therefore
re) }
{pa
+ qb + re)}
+ r 2 (c (_/; + qb + ?*c)j
.
two quantities vanish, by virtue of equations
and [9]; and therefore
this in the first
(n
2
)
=p. 2
denominator of equation
[8]
Substituting
(an).
[1],
the equation
becomes
("i~2 Kl i) & n i
or
(#&,
+ 7'oJ
r
or
Bn 1
+ (na ~~P at)
+ (qb + re)
q
iV(bSn.l
'
1+ r (cjw,
b9Sn92
'
c8s
^n2
Sna
+ &C.))
,
+ &c.)J
+ ^ c = 0>
4 &c. =

[10];
DETERMINATION OF SEVERAL ELEMENTS.
which
under the new assumptions, the equivalent of
is,
equation
[1],
and on the truth of which
validity of the
is
true
one
Now
new assumptions.
hand
for its left
=0
by equation
consists of
[2],
taneously satisfied
and
will
two
parts, of
the
which
by equation
[1], [2], [3],
therefore
Article 91 gives the values of x,
depend the
the equation [10]
and the other
Consequently, the equations
[3].
7/
are simul
assumption of
whose probable error
is
minimum.
If
93.
we
investigate,
by a
similar method, the as
sumption which will give
error
is
for y the value whose probable
minimum, we have only to remark that the equa
tions [4], [5], [6], are symmetrical with respect to x, y,
and
and therefore when treated
z,
manner
as for
a;;
that
is,
the probable error of
of the same
minimum.
z,
and
is
z,
minimum.
in the
same
result for
In the same manner,
as determined from the solution
equations, treated in the
The problem,
same
the probable error of y, as determined
from their solution,
y,
for
as for x, they will exhibit the
same manner,
is
therefore, of determining values of x,
to satisfy, with the smallest probable error of x, y,
z,
the numerous equations
axx
+ l y + c z =fv
aax
+ b2y + czz =/s
&c.
ADVANTAGEOUS COMBINATION OF MEASUEES.
78
is
completely solved by solution of the equations
x 2
.
a?
(a
(aft)
{ah)
iac)
=2
(a/),
+y 2
(//)
+2.2
{bo)
=2
(*/),
(Jc)
=2
(</).
+y 2
(ac)
Suppose
94.
+y 2
.
2
(
that, instead of proposing to ourselves
the condition that the probable errors of the deduced
values of
x, y, z, shall
condition
be minimum, we had proposed this
sum
that the
maining after correction
and
z,
of the squares of the errors refor the
deduced values of
x, y,
or
^.{ax + hj + czfy,
shall
On
be minimum.
with respect to
each equation
differentiating
and taking their sum, we should have
x,
obtained
2
and similarly
[a
for
{ax
y and
+by + cz /)] =
Z{b.{ax + by + czf)}=0,
%{c.{ax+hj + czf)}=0;
the very same equations as those found above.
In conse
quence of this property of the equations, of giving such
values of
x, y,
and
z,
that the
sum
of squares of errors
remaining after their application shall be minimum, the
method
is
squares."
very frequently called " the method of
This term
is
very unfortunate
it
minimum
has
fre
quently led investigators to suppose that the subject of
MINIMUM SQUARES OF RESIDUAL ERRORS.
the
minimum
presented
is
sum
the
whereas
discordances,
it
when
79
of squares of discordances as first
ought to be the sum of squares of
have the same
so multiplied as to
probable error.
It
95.
is
easy to see that the same principles apply,
the same remarks hold, and the same result
when
the
number
of
restricted to three,
is
unknown
obtained,
is
elements, instead of being
any whatever. The rule
is
universal;
multiply every equation by such a factor that the probable error of the righthand term will be the same for
all
multiply every altered equation by
of one
unknown
equation
and
so
on
the same for the
for every
coefficient
its
sum
second unknown
quantity, and take the
unknown quantity
for a
new
quantity
and thus a num
ber of equations will be found equal to the number of
unknown
quantities.
In order to exhibit the probable error of x thus
96.
determined,
cess.
It
we may proceed by
will
a purely algebraical pro
however soon be found that
results of intolerable complexity.
We
it
leads to
would recommend
the reader to introduce numbers as soon as possible for
every symbol except
all errors spring).
(that quantity
from whose error
In the following explanation, however,
of the succession of steps, the reader will easily see to
what extent he can advantageously
retain the symbols.
It is first necessary to find the factors of the equations
[4], [5], [6], of
Article 91, or the last equations of Article
SO
93,
ADVANTAGEOUS COMBINATION OF MEASURES.
which
will eliminate
y and
They
z.
are easily found
to be,
For equation
[4],
S.^xS.c ^. he)'.
For equation
[5],
For equation
[6],
There
number
trouble
no
is
ac x
& x 2 &c
.
ab x
ac
"Z
c\
x2. b".
quantities exceeds three
so great that
is
S be 2
when the
difficulty in finding the factors
unknown
of
it
will always
but the
be best to use
numbers.
Applying
these,
we
obtain
x = P.%(af)
where P,
+ Q.^(bf) + R.t(cf)
Q, B, are numbers, but af, bf,
are for the
cf,
present retained in the symbolical form.
Now
we examine the form
if
quantities
f f
lt
in
collect together all the multiples of
tity,
we
(pfli
+ Qh + Bc^f +
t
We have
here a
number
(Pa,
Qb,
we
+ Bc& f + &c.
Remarking that the probable
r.oixj
(Pa,
is
of independent fallible quan
which the formula of Article 52
quantities f^f^, &c.
/p.
if
each individual quan
shall find
tities, to
apply.
which the individual
&c. enter into this expression, and
will properly
error of each of the
supposed to =e, we obtain
+ Q\ + Re,) 3 +
= z.(ra + Qb + iicy;
(Pa,
+ QK + PcJ + &c.
SMALL VARIATION OF THE FACTORS
may
which
but which
by an
will
be very far easier to evaluate in numbers
entirely numerical process.
would be exactly
The
97.
for finding the probable errors of
relaxation of the rules for determining the
which in reference
is
By taking advantage of
may sometimes be materially
In actual applications, the numbers at 2 &c.
,
In prac
desirable accuracy will be secured for the result,
striking
off,
in the factors only, all the latter decimals,
leaving only one or two significant figures.
different factors will
produce different
certainly
The use
results,
of
but not
more inaccurate results we have no means of
knowing which are the best we only know that,
necessarily
if
and
quantities,
&c, usually consist of troublesome decimals.
tice, all
by
unknown
theoretical reason.
this relaxation, the labour
diminished.
is
admissible also in the treatment
of equations applying to several
same
factors of the equations,
to the treatment of simple measures
explained in Article 73,
for the
y and
similar.
most advantageous values of the
b 1}
PERMITTED. 81
be exhibited in symbols of great complexity,
it
The operation
s
IS
we repeat the
process in an infinity of instances, the fac
tors corresponding accurately to
with results whose errors
are,
minimum
will furnish us
on the whole, a
than those originating from other
factors.
little
smaller
ADVANTAGEOUS COMBINATION OF MEASURES.
82
Instances of the formation of equations applying to
14.
several nnknoivn quantities.
It will perhaps
98.
tions,
we
bo instructive to shew how equa
such as those treated above,
will take
two instances
For this purpose,
arise.
one of very simple and one of
very complicated character.
Instance
99.
It is required to
1.
determine the most
probable values of the personal
equations between a
number
C,
of transit observers
A, B,
D, &c; where the
observers have been brought into comparison in
many
combinations, or perhaps in every possible combination
but never more than two at a time.
Use the symbol
sons between
(ab) to
denote the number of compari
and B, and
AB
for the symbolical
value of the personal equation between
parisons.
And
and B, (AB)
suppose that the probable error of each
single comparison
is
numerical value deduced from the mean of com
for its
is e.
Therefore
Then the probable
error of
(AB)
when we have formed the equation
AB = (AB),
in
which the
e
,
/(ab)
last
we must,
term
srm
is
liable to
tc the probable
error
in conformity with the recommendation
in Article 87, multiply the equation
by
*J(ab),
and then
INSTANCE: PERSONAL EQUATIONS.
its
probable error will be
e.
Thus we
find,
83
for the dif
ferent comparisons, the following equations, all liable to the
same probable
error e
V(aj)
V(c)
A  s/(ab) B = sjiab)
A  *J(ac) G = V(c)
(AB),
(A C),
(J5C),
&c.
Vt&c)
B*J(bc)
C= V(6c)
&c.,
and these equations are exactly such as those
88,
though in an imperfect form.
tions
must therefore be formed by the
Thus we
The
find
first
in Article*
The determining equarule of Article 93.:
equation
be formed by the
is to
sum
of the
following,
(ah)
(ac)
A  (ab) .B = + (ab)
A  (ac) C = + (ac)
.
(AB),
(A C),
&c.
The second equation
is
to be
formed by the sum of
the following,
(ab)
B  (ab) .A = (ab).
(bc).B(bc).
(AB),
C= + (bc).(BC),
&c.
F 2
ADVANTAGEOUS COMBINATION OF MEASURES.
84
Thus we obtain the simple
rule
Form each equation for comparison of two
taking the mean of all their comparisons.
Multiply each such equation by
its
observers
number
by
of compari
sum
of all the
individual comparisons of those two observers
had been
This
sons.
is,
in fact, the
same
as if the
taken.
In the various multiplied equations which contain
make
the coefficient of
changing
in every equation positive (by
the signs of the equation where necessary),
all
and then add
all
together to form a determining equation.
In the various multiplied equations which contain B,
including,
make
if
necessary, one from the lastmentioned series
the coefficient of
changing signs
if
in every equation positive (by
necessary),
and add
all
together to form
a determining equation.
In the various multiplied equations which contain
including,
make
if
necessary, one from each of the last
the coefficient of
changing signs
if
two
C,
series
in every equation positive (by
necessary),
and add
all
together to form
a determining equation.
And
so
It will
may be
through
all
the observers.
be found that one of the determining equations
produced by a combination of
and therefore
all
the other de
termining equations
assume a value
one of the quantities, A, or B, or
C, &c.
for
it
is
necessary to
INSTANCE: GEODETIC TRIANGULATION.
100.
Instance
85
In a net (not necessarily a simple
2.
chain) of geodetic triangles
one or more sides have been
actually measured, or so determined by immediate refer
may
ence to measured bases that they
measured; in some of the
be considered as
triangles, three angles
been measured, in others only two
at
some
have
stations, all
the angles round the circle have been observed, at others
not
all
at
observed
some
it
stations, astronomical
required
is
to lay
azimuths have been
down the
rules for de
termination of the positions of the different stations.
101.
It
is
necessary to determine the value of
first
And
probable error in each of the observations.
this is
not to be done by a simple rule, because the observa
For instance, the horizontal angle
tions are not all alike.
between two signals
is
liable to
error from
(1)
instrument, (2) error of pointing to one signal,
of pointing to the other signal
error
is
is
easily formed.
But
for the
of a given signal, the sources of error are,
instrument,
(2)
bable error of this last
The
azimuth
(1) error of
error of pointing to the signal, (3) error
of pointing in the direction of the meridian
mate
error
and when each probable
ascertained, the probable error of horizontal angle
between signals
others.
error of
(3)
may
and the pro
be very different from the
linear measures will require a peculiar esti
of probable error.
All must, however, antecedent to
all
other treatment,
be so found that the probable error of every measure, of
whatever kind, can be
specified.
ADVANTAGEOUS COMBINATION OF MEASURES.
86
The next
102.
step in this (and in all other compli
cated cases) will be, to assume that every coordinate of
which we are seeking,
station
and numerically expressed.
so small in scale that
we may assume,
its
is
Thus,
area
approximately known,
if
may
the triangulation
is
be supposed a plane,
two rectangular coordinates of
for the
every point, numerical values, each subject to a small
correction (which corrections
it is
the object of the whole
If the triangulation
investigation to ascertain).
is
in scale that the spheroidal form of the earth
we may assume,
regarded,
and longitude of every
for
so large
must be
the astronomical latitude
point, numerical values, each sub
ject to a small correction.
With
103.
these numerical values and symbolical cor
rections, every fact
must be computed
which has been the subject of measure
and the computationresult must con
of two parts, one numerical,
sist
area
is
For
and the other multiply
Thus; suppose that the
ing the symbolical corrections.
plane, and that the rectangular coordinates are
1st station,
For 2d
For 3d
station,
station,
o^
+ Ba^ b^&b^
or
+Sa.,,
b,_,
bs
o3
Ba8
+ 8b^,
+
Bb&
&c.
(where a
Ba,
&c, Bb lf
direction of a
&c,
Bb, &c.
is
l\,
\, &c, are numerical, and Ba
Suppose that the
t
are symbols only).
parallel to the meridian.
of the second station as viewed from the
Then the azimuth
first is
instance: geodetic triangulation.
( 2
+(hK)
x)
'
(7
* ,
and where
merical, except Bat , 8l\, Ba2
quantities are nu
all
For convenience we
Bb,
lJ
faad +(Kh)
where tan
87
will
write this,
True azimuth of 2d station as seen from
=
And
a + a,
Ba
*h A*
>
1st station
K  A,
iu like manner,
True azimuth of 3d station
=C +A
3
lfi
8a 3
A
lfi
as seen from 1st station
Ba,
+ B h3
cb3
l3
Bb t
Now if the azimuth of the 2d station had been observed
at the 1st,
and found <y2 (subject
then the comparison of the
first
to error of observation),
formula with this would
give the equation
A*
If,
K  A*
Sf'i
B*
Bh *
~B
i,*
hh i
= v*  G*'
however, no azimuth from the meridian had been
observed, but only the horizontal angle #23 between the
2d and 3d
stations,
which
is
the same as
subject to errors of observation
A*M 
VV
<y 3
%, and
then we should have
J '.s A J ^ + *&* ~ B^h
is
ADVANTAGEOUS COMBINATION OF MEASURES.
88
The
distance of the 2d station from the
= V[K
a
+ (^
1)
vKa,a>+(Kw (K
first
W
~ Sai)
which may be written
A+M
1A
Now
if
 SaJ + N,t (BK 
(Saa
the distance from the
had been measured, and found
Bb,).
station to the second
first
=X
subject to error of
observation; then the comparison with this formula would
give
lfi
lfi
+ NUi
Ba x
8ba
 X12
Each of these equations
104.
hand, a
M
ca2
fallible,
quantity
the
first
S& x
=\ L
contains, on the right
contains
<y a ,
the second
The probable error
of each of these, as we have said in Article 101, must be
supposed known and then, the equation must be divided
contains #23
the third contains
X.,.
by a
divisor proportional to that probable error.
operation being effected,
tions
whose probable
we
errors are all equal
of Article 93 can be applied
and we
of determining equations equal in
of
unknown
stations.
have a
shall
quantities
that
is,
shall
number
to
scries of
This
equa
and the rule
have a
series
number
double the number of
to the
GENERALITY OF THE THEORY.
The reader who
105.
will well
.89
examine such an
in
stance as this will be struck with the perfect generality
and great beauty
He
of the theory.
will see that,
what
ever has been observed, and in whatever shape, as a fact
of observation, will enter with
its
proper weight into the
He may
formation of the final determining equations.
exercise his fancy in introducing different circumstances.
Suppose, for instance, that a referringsignal has been
The
used.
ringsignal
it is
case
assumed azimuth of that
be a new unknown quantity.
v/ill
combined only with observations of
it
produces one form of equation
combined
it
correction of
also
refer
Sometimes
signals, in
which
sometimes
it
is
with meridianobservations, in which case
produces another form of equation.
of observations,
it
may
In some batches
be necessary to use the theory of
"entangled measures" (see Articles 74 to 80) before the
probable errors can be properly found.
sure
is
Whatever mea
made, a proper corresponding equation can be
formed, and the proper cautions accompanying
it will
soon
present themselves.
10G.
It
may
occur to the reader as a difficulty, that
the quantities concerned are not homogeneous
sure of
0,
for instance,
units of length.
measure
is
the mea
This, however,
is
only apparent.
Any
expressed by means of certain units, and
probable error
we
being in angle, and that of X in
is
expressed by the same units
and,
divide each equation by a divisor proportional to
probable error,
we do
in fact
its
when
its
produce an equation of abs
ERRORS OF OBSERVATIONS REQUIRED
00
numbers, whose probable error also
tract
The whole are then
number.
is
an abstract
comparable, in
strictly
whatever kinds of measure they have originated.
The
107.
tions
solution of so
numerous a
of course troublesome.
is
It
is,
series of equa
however, no more
troublesome than the nature of the subject strictly re
And
quires.
to
it is
be considered that
it
gives to every
observation of every kind exactly the weight that
to
it
that
it
system of results
is
the
and that that
the most probable.
Treatment of Observations in which
that
due
leads to one distinct system of results, with
out leaving any opening for uncertainty
15.
is
it is
required
Errors of Observations rigorously satisfy
some assigned conditions.
108.
of
In the equations considered in Article 87, each
which gave the
tors,
the
unknown
effects of
combining, with different
that observations might thereby be better represented,
is
to
fac
corrections of certain elements, in order
it
be remarked that there was no expectation that the
result
of combination of these corrections of elements
would exactly represent the observations, or that any exact
relation could be assigned as
rections
existing
among
the cor
which were to be found; or between "the result of
applying those corrections," and "observations".
109.
But there
are instances in which the nature of
the problem requires that,
among
the corrections which
/
TO SATISFY ASSIGNED CONDITIONS.
91
are to be found, a prescribed equation shall be rigorously
The nature and treatment
maintained.
of these will best
be understood from examples.
110.
Instance
In a geodetic triangle, observa
(1).
On comparing
tions of the three angles are
made.
sum with the quantity
that sum ought to be
+ spherical
180
equal,
it
How
requires the correction A.
is
excess, to
their
which
found that the sum
ought that correction to
be divided among the three angles, their probable errors
known
of observation being
111.
Ev E,v E
Let
be the three corrections required,
the probable errors of observation being et
the
first
consideration
E
E
is
e2
es .
Then
is,
to
be as small as possible,
here corresponds to the quantity
e[
The quantity
ax\by\ cz
the latter being the residual difference
in Article 87;
between an observed quantity and a quantity computed
from corrected elements, which difference is to be made
"as small as possible."
be exhibited in the
Hence we may
The
equations are therefore to
same form.
state the equations thus
E=
E = 0,
with probable error e v
tf.>fc
0,
e2
ERROKS OF OBSERVATIONS REQUIRED
92
wo stopped at this form, we could not obtain a valid
the number of equations being tlie same as the
number of unknown quantities, in which case no solution
If
solution
depending on probabilities can be obtained.
Now we
112.
and use
it
introduce the condition
one of the quantities, as
to eliminate
Then
the equations become,
i^=0, with
probable error
ex
= 0,
e,,
AE^E^O,
es .
E.z
Here we have three equations to determine two quantities, and the process of Article 93 may be followed.,
113.
Dividing by the probable errors, we have these
equations, in each of which the probable error
=1
!
e3
e3
e3
and therefore, by the process of Article 93, forming a
final
equation principally for
Ev
by multiplying each
93
TO SATISFY ASSIGNED CONDITIONS.
equation here by
its.
coefficient of
and adding the pro
ducts,
{ey
(?/
Similarly, forming a final equation principally for E,,
by multiplying each equation by
its coefficient
of
z,
and
adding the products,
WWW
&
Comparing these two equations,
at
but
once infer from this that
*it
may be more
completely.
E^ = E
r
W e might
h has the same
value,
satisfactory to solve the equations
Eliminating
from the
first
the relation just fouud,
Therefore
^^.jgr^q^ii
whence, by the relation found,
w+w+w
equation, by
ERRORS OF OBSERVATIONS REQUIRED
1)4*
and by subtracting their sum from A,
EA
'
^
\2
W+W' + to"
Hence, the corrections to be assigned to the different
angles ought to be proportional to the squares of their
respective probable errors.
Instance
114.
nals can be seen
From
(2).
a theodolite station, n sig
the angles, between each signal and the
next in azimuth, are independently observed
which ought to be
rection
360, is 360
B to be divided
= 0,
their sum,
ought the cor
The equations
115.
 B how
in this instance will
with probable error
e1
E,= 0,
e2 ,
K=
en .
0,
be
Then, by the equation
E^E.^&c. + E^B,
the last of the equations
is
changed into
B E E &c En_ = 0,
1
with probable error
en
and the equations are to be treated in the same manner as
in Instance (1)
and a similar
result is obtained
namely,
TO SATISFY ASSIGNED CONDITIONS.
that
the
corrections
95
be assigned to the different
to
angles must be proportional to the squares, of their respective probable errors.
The next
116.
instance will be
Instance
(3).
more complicated.
In the survey of a chain of
tri
angles, a hexagonal combination of the following kind
occurs, in
all
to
which every angle
are liable to error
is
observed independently;
to find the correction
which ought
be made to each.
Let the angles be denoted by the simple numbers
let
their corrections
sought be
probable errors of observation
(1),
[1],
[2],
(2), &c.
&c, and their
ERRORS OF OBSERVATIONS REQUIRED
06
Then we have the
= 0,
[2] =
[1]
equations,
with probable error
(2);
[18]
And now
unknown
(1),
= 0,
(18).
have to consider how many of these
Ave
quantities can be eliminated
by virtue of the
geometrical relations.
Adding the angles
117.
comparing the sum with
[1]
[2]
Then
[3]
at the central station,
+ [4] + [5] +
[G]
= a known
quantity A:
in the six triangles,
[1]
known quantity
+ [9] + [10]
[3] + [H] + [12]
[4] + [13] + [14]
[5] + [15] + [10]
[6] + [l7] + [18]
When
we
D,
E,
F,
G.
corrections satisfying these equations are apshall
have a set of six triangles, with angles
consistent in each triangle
that they
B,
C,
[2]
plied,
and
300,
fill
up 360
and which
so adhere together
at the central station
nevertheless
TO SATISFY ASSIGNED CONDITIONS.
it
might happen
d from
c,
from
that, in calculating b
d,
from
find a value a' differing
a'
from
But
a.
be found rigorously equal to
it is
a, c
f,
from
b,
we should
necessary that
Tracing the calcu
a.
through the several triangles,
lations
from
and a from
e,
97
it is
found that this
equation gives (with corrected angles),
sin
sin 9
sin 8
sin
10
sin 11
sin
13
sin
sin 12
sin
14
sin 16
15
sin
17
sin 18
and, taking the logarithms, with the addition of symbols
for the corrections,
log sin 7
.
log
sin 8
+ log
sin 9
log
sin
10
+ log
sin 11
log.sinl2 + log.sinl3 log.sinl4 + log.sinl5 log.sinlG
f log sin 17 log sin 18
.
+ cot7
x[7]
cotSx
[8]
+ cot9x [9]&c
+ cotl7x [l7]cotl8
=
We
x [18]
0.
shall use the
this expression,
which
symbol
is
to denote the first part of
known
quantity.
Thus we have eight equations
By means
fied.
quantities
tities to
118.
[1],
of these,
[2],
we
be rigorously
to
satis
are to eliminate eight of the
&c, and there will remain ten quan
be determined by eighteen equations.
Suppose
corrections [1], [2],
for instance
&c, as
[2]=
we
decide to eliminate the
far as [8].
We
have
C [9] [10],
[3]=D[11][12],
A.
ERRORS OF OBSERVATIONS REQUIRED
98
[4]
='[13] [14],
= ^[15] [16],
[6] =[17] [18].
[5]
Substituting these iu the
equation,
first
[\]=ACDEFG
+ [9] + [10] + [H]+[12]+[13]+[14]+[15]+[16]+[l7]+[18].
Then
[7]
= [l][8]
=A+B+C+D+E+F+G
[8][9][10][ll][12][13][14][15]
_[1G]_[17][18].
Substituting this in the last equation of Article 117,
L + cot 7 x
(
A + B + C+ D + E+ F+ G)
cot7x{[8]+[9]+[10]+[ll]+[12] + [13]+[14]+[15]
 cot 8
[8]
cot 12 x
+ cot
15 x
cot 18 x
From
+ cot 9
+ cot
[15]  cot
[12]
[9]
 cot
+ [17]
10 x [10]
+[18]}
+ cot 11
x [11]
13 x [13] cot 14 x [14]
1G x [16]
+ cot 17
x [17]
[18].
this equation, [8]
&c. as far as [18].
expression, [7]
as [18].
[16]
Thus
is
And
is
found in terms of
found in terms of
all
pressed in terms of
the corrections
[9], [10]... [18],
[9], [10],
it
in the preceding
[9],
[10], &c. as far
substituting
[1],
[2], ...[8],
are ex
TO SATISFY ASSIGNED CONDITIONS.
The primary equations
119.
[1]
of probabilities are,
with probable error
[2]=0
(1),
(2),
[18]=0
Of
99
(18),
now be changed
these, the first eight will
into the
following
[9]
[10]+...
+ [18] = vl+
C+D + E + F+G,
with probable error
(1)
C,
(2)
[11]
+ [12]=2>,
(3 )
[13]
+[14]=^,
(4)
[15]
+ [16]=^
(5)
[17]
[9]
("series
[10]
[18]
of multiples]
G,
r
(6)
series of
]
known
of[9],[10]...[18],U
[
expressing [7]
series of multiples')
(7),
(quantities!
r
series of
]
i of[9],[10]...[18],l = i known
expressing [8]
The remaining equations
[9]
= 0,
(8).
[quantities'
will retain their simple form,
with probable error
(9),
[10] = 0,
(10),
= 0,
(18).
[18]
G 2
ERRORS OF OBSERVATIONS REQUIRED
100
Each
120.
be divided by
of these eighteen equations
is
then to
probable error, and we thus obtain the
its
following equations, in each of which the probable error
= i;
[9]
(i)
[10]
[18]
_ A + C + D+E+F+G
"
(1)
with probable error
W
and so through
The
'
(2)'
(2)
1.
the equations.
all
equations so divided, having the same probable
error, are in
fit
state for application of the
method
of
Article 93.
The
for [9]) will
be formed by multiplying each equation by
the coefficient of
ducts
first
[9] in
of the final equations (principally
that equation, and adding the pro
the second of the final equations (principally for
[10]) will
be formed by multiplying each equation by the
coefficient of [10] in that equation,
ducts; and so on to [18],
the values of
[9],
From
and adding the pro
the equations thus formed,
[10].. .[18], are
found; and by substi
tuting these in the formula? of Article 118, the values of
[1], [2]... [8]
121.
are found.
It is particularly to
be observed that, although
in the changed equations of probabilities
such quantities as
[1],
[2],
corresponding probable errors
must be
left in its place.
error will be
remarked
we
eliminate
&c, we do not eliminate their
(1),
(2),
&c, each of which
This retention of the probable
in Instances (1)
and
(2).
TO SATISFY ASSIGNED CONDITIONS.
122.
The complete
solution
would scarcely ever be used in
is
101
so troublesome that
practice.
it
Probably some
process like the following would be employed, with sufficient
accuracy
Divide the error
by the process of Instance
(2),
and
use the corrected angles in the process that follows.
Divide the errors B, C,...G, by the process of Instance
(1),
and use the corrected angles in the process that
fol
lows.
Apply the
last
equation of Article 117, by a process
nearly similar to that for A.
Repeat the process
for dividing
A'
(the discordance at
the center produced by the angles as last corrected).
Repeat the process
for dividing B',
continue this operation as often as
may be
...
G'.
necessary.
And
PART
IV.
ON 3IIXED EKKOES OF DIFFERENT CLASSES, AND
CONSTANT ERRORS.
Consideration of the circumstances under which the
16.
existence of
Mixed Errors of Different Classes may
and investigation of their separate
he recognized,
values.
When
123.
made, day
which
is
successive
after day, of the
either
series
of
observations
are
same measurable quantity,
invariable (as the angular distance be
tween two components of a double star) or admits of being
reduced by calculation to an invariable quantity (as the
and when every
known instrumental correction has been applied (as for
still
zero, for effect of temperature upon the scale, &c.)
apparent angular diameter of a planet)
it will
sometimes be found that the
one day
differs
result
obtained on
from the result obtained on another day
by a larger quantity than could have been anticipated.
The idea then presents itself, that possibly there has been
on some one day, or on every day, some cause, special to
the day, which has produced a Constant Error in the
measures of that day.
the evidence
error.
for,
It
is
our business
and the treatment
of,
now
to consider
such constant
ADMISSIBILITY OF CONSTANT ERROR.
The
124.
103
existence of a daily constant error, that
is,
of an additional error which follows a different law from
ought not to be lightly assumed.
the ordinary error,
When
made on
observations are
only two or three days,
and the number of observations on each day
tremely great, the mere
fact, of
discordance from day to day,
The
constant error.
to a
"run
bable.
not ex
accordance on each day and
is
not sufficient to prove a
existence of an accordance analogous
of luck" in ordinary chances
If this
is
is sufficiently
pro
be accepted, as applying to each day, the
whole of the observations on the different days must be
aggregated as one
subject to the usual law of error.
series,
More extensive experience, however, may give greater confidence to the assumption of constant errors and then the
;
treatment of which
we proceed
to speak will properly
apply.
First, it ought, in general, to
125.
there
is
possibility of error, constant
ing from day to day.
Suppose, for instance, that the
distance of two near stars
is
observed with some double
image instrument by the method
alternately right and
be established that
on one day but vary
left.
of three equal distances,
It does not
appear that any
atmospherical or personal circumstance can produce a constant error; and, unless
like those to be
entertain
it.
we
are driven to
it
by considerations
mentioned in Article 129, we must not
But suppose, on the other hand, that we have
measured the apparent diameter of Jupiter.
It is evident
that both atmospheric and personal circumstances
may
MIXED ERRORS, AND CONSTANT ERRORS.
104
and here we may admit the
sensibly alter the measure;
possibility of the error.
Now
126.
separately, and,
us take the observations of each day
let
by the rules of Articles GO and
Gl, investi
gate from each separate day the probable error of a single
We may
measure.
expect to find different values
(the
mere paucity of observations will sufficiently explain the
difference); but as the individual observations on the dif
ferent days either are equally good, or (as well as
we can
judge) have such a difference of merit that we can approxi
mately assign the proportion of their probable
errors,
we
can define the value of probable error for observations of
standard quality as determined from the observations of
we must then combine
each day;
these,
with greater
weight for the deductions from the more numerous observations,
and we
shall
have a
final
value of probable error
of each individual observation, not containing the effects
of Constant Error.
infer
Article
55,
Eesult
;" still
The
"
From
this
we
can,
by the
rule
of
the "Probable Error of Each Day's
not containing the effects of Constant Error.
Result of Each Day," also not containing any cor
rection for Constant Error,
is
given by the
mean
of deter
minations for each day.
We
must now attach to the numerical value of
"Result of Each Day" a symbol for "Actual Error of
Result of Each Day;" and take the mean of all these
127.
compound
days
quantities, numerical
and symbolical,
for all
the
(the combinationweights being proportional to the
"
PROCESS FOR DISCOVERY OF CONSTANT ERROR.
number
of observations on each day, unless
any modifying
circumstance require a different proportion).
may
is
The
be regarded as "Final Result."
to
"
105
mean
This
Final Result"
be subtracted from the "Result of Each Day;" the
remainder
is
the "Discordance of Each Day's Result."
For each day
it
consists of
series of multiples
of
two parts; a number, and a
the symbols for " Actual Error of
all
Result of Each Day."
128.
Now
the Discordance
treat
number accompanied with multiples
itself
an Error, and investigate the
by the
(consisting
of the
of symbols) as being
"
Mean
Discordance
rule of Article 26 or 59; a value of
"Mean
Dis
cordance" will thus be obtained, consisting of a number
accompanied with a
"Actual Error."
an independent
series
of multiples of symbols
of
Consider each day's "Actual Error" as
fallible
quantity whose Probable Error
is
that obtained in Article 126, and form the "Probable
Mean Discordance" by the rule of Article 52.
Thus we have, for Mean Discordance, a formula consisting
Error of
of
two
parts,
(1)
(2)
A
A
namely,
numerical value.
number expressing the probable
error in the
determination of that numerical value.
129.
And now
it
will rest entirely in the
judgment of
the computer to determine whether the simple numerical
value
(1)
or not.
just found,
is
to
be adopted
for
Mean Discordance
It is quite clear that, if (2) exceeds (1), there is
no
MIXED ERRORS, AND CONSTANT ERRORS.
106
sufficient justification for the
that
it
is,
assumption of a Discordance,
of a Constant Error.
If (2)
is
much
less
than
(1),
appears equally clear that a Constant Error must be
assumed
and
to exist,
adapted for
Mean
(1)
or
any value near
or Probable Constant Error, will be found
this
by 08153,
may he
by multiplying
as in Article 31.
The reader must not be
130.
it
The Probable Discordance,
Discordance.
startled at our referring
these decisions to his judgment, without material assist
The Calculus is, after all, a
of the mind are worked
but which must be directed by the
ance from the Calculus.
mere
tool
by which the decisions
out with accuracy,
mind.
In deciding on the admissibility of Constant Error,
after giving full
it will still
weight to the considerations of Article 129,
be impossible, and would be wrong, to exclude
the considerations of Article
125,
and these cannot be
brought under algebraical or numerical
131.
These investigations
rule.
suppose that
the
cordance of Each Day's Result" cannot, so far as we
"Dis
know
antecedently, be referred to any distinct assignable cause.
But
if
there
should appear to be any such cause,
for instance, if
we
as,
conceive that the observations of one
person always give a greater measure than the observations
of another person,
it
will
be easy to apply an investigation,
analogous to that just given.
The observations
of each
person should be separated from those of other persons and
collected together;
from the collected group of each per
son's observations, a
Mean
Result and Probable Error of
DISTINCT CAUSE OF CONSTANT ERROR.
Mean
Result for each person must be found
107
and then the
reader must judge whether, in view of the amount of Pro
bable Errors, a Personal Difference of Results
The
or required.
investigation
is
ing, in this respect, that it arrives at
Difference
And
of Results,
the result
is
admissible
is
simpler than the preced
a Simple Personal
and not at a Mean Discordance.
simpler than the
last,
because
it is
a Con
stant Correction to the results of one person, instead of an
uncertain correction liable to the laws of chance.
Treatment of observations when the values of ProError for different groups, and probable
17.
bable Constant
error of observation of individual measures within each
group, are assumed as known.
132.
tions
When numerous
and extensive
series of observa
have been made, as in Articles 126, &c,
sufficient to
determine the Probable Value of the socalled Constant
Error (which
is
in fact
an Error varying from group to
group) and the ordinary probable error of an individual
observation in each group
suppose that there are
occasional observations, in limited groups, for which
desirable to define the rides of combination.
justified, for
We
made
it is
are not
each of these limited groups, in assuming a
value for the Constant Error, or Variable Error of the
Second
as
Class, applicable to that
group; we must treat
it
an uncertain quantity, and ascertain the combination
weights, and the probable error and theoretical weight of
final result,
under the
effects of the errors of the
two
classes,
MIXED ERRORS AND CONSTANT ERRORS.
108
by an operation analogous
the errors are only of one
In the
133.
first
to those
which are applied when
class.
group of observations,
the actual
let
value of the error of second class be ^G) in the second
group
in the third
of each being
group
And
c.
values of the errors of
being
And
e.
let
cessive groups be
be
1}
z
z
x %, t 3
E E
X ,
jn,
&c;
n,
z
2 v
the probable value
Ev E E
&c.
number
the
C, &c.
class (or ordinary errors) for the
fir^st
successive observations be
the second group
in the first group, let the actual
3>
&c.
for those in
of observations in the suc
Let the combination factors
&c.
z
2 2
the probable value of each
z
2 3
&c.
xt
z2
z3
&c.
and
so for successive groups.
Then
the actual errors of the separate measures will be
&c.
&c.
INVESTIGATION OF COMBINATION WEIGHTS.
And
109
the actual error of the final result, obtained by
combining the separate measures with the combinationweights above given, will be the fraction, whose numerator
is
(a +
A+ A + &c
+ (A.
)
i#+ (a + a + &c
^ + AA + AA + &c.)
+ <aA+aA+&.)
+ (A>*^i + :* + **) + *
and whose denominator
is
(A + A + A + &C + (A + 2* + &c + (A + 3 ? + &c) +&c.
2
The square
134.
result,
of the probable error of the final
found in exactly the same way as in
ing cases, will be the fraction whose numerator
(A)
(A)
+ & c.}
and whose denominator
{(A
+ A + &c + (A +
This
is
to
+ {(^Y +
z
s
2>
making each
is
*2
+ &c + (A + s* + &c + &c.}
)
)
2
.
be made minimum with respect to the varia
&c &c

+ &c.}
preced
is
tion of each of the quantities
Z
9 i>
/2 )
(,
all
sx
z# &c,
Differentiating with
differential coefficient
in former instances,
we
= 0,
z v z 2
&c,
respect to each,
and treating as
find successively,
an indeterminate constant),
(putting
for
'
MIXED ERRORS AND CONSTANT ERRORS.
110
zx
First,
therefore, for each of these
Second,
n. x z.c
n. 2z.c
n. 3 z.c
z%
= &c.
~3
we may use
z.e
= A,
= A,
= A,
z.e
z.e
ii
the symbol
z.
&c.
from which we obtain
_A_
xZ
which
n
x
+e
2 '
applicable to every observation in the
is
first
group
A
which
group
applicable to every observation in the second
is
;
and
so
on through
In the numerator of expression
135.
probable error of result,
so for other groups,
y /
.
the groups.
all
2
.
=A
c~
+ &c. +
(x
square of
for the
&c, we
insert
+ ji
2
.
and
+ &c.
+ &c),
and the same substitution converts the denominator
du
+ ^c
2
)
and the square of probable error of result
A
n
x
z,
becomes
it
2
.
z
z
x x, x 2
if for
z
x
ji
+ &c.
to
DISTINCT CAUSE OF CONSTANT ERROR.
which with the values of
z,
z,
Ill
&c. found above, becomes
1
77
Vt5 + &C.
Or
(probable error of result)
136.
If,
as in Article 131,
upon some
fix
group,
Error,
we
distinct cause of
\
conceive that
the Constant
that group of results
may
then be reduced
For the determination
of the probable error of the result of the
it
group so cor
must be borne in mind that the determination
of the Constant
to
is
Let A, B,
liable to error.
n terms, be the actual
errors,
and
a, b,
c,
C,
responding to that in which
we
assuming a Constant Error.
D, &c.
d, &c.
probable errors of the means of various groups,
for
we can
Constant Error for one
all the others being assumed free from Constant
aud can ascertain with confidence the amount of
by application of the Constant.
rected,
the
cor
suspect sufficient reason
The
actual error of de
termination of Constant Error will be
+ &C.
AJ3+C+D
ii
and the probable
will
error of determination of Constant Error
be
V?
b~
+ <y + d~ +
(nlf
&c.
112
CONCLUSION.
But the
corrected result of that group will be liable to the
actual error
_B+ C+D + &C.)
n
and
its
B+G+D + &C.
probable error will be
+ c + d* + &c. \
2
ft*
In
fact,
sults,
by referring that
and
(niy
result to the
so determining its correction,
mean of other rewe entirely deprive
that result of any original value in the application of these
groups.
But
if,
on another occasion, there were observations
made by the same person or under the same circumstances
as the observations A, then the determination of Constant
Error and of
its
probable error just found would be pro
perly applicable.
These
conclusions will be varied according to the
various assumptions
ficulty in
made
the reader will have
little dif
applying the theory of preceding Articles to any
of them.
CONCLUSION.
137.
In the practical applications of the Theory of
Errors of Observations and of the Combination of Observations which have fallen under our notice, the following
are the principal sources of error and inconvenience.
ERRORS TO BE AVOIDED.
113
In some instances, measures have been combined
(1)
by a method
minimum
of "
squares " without reference to
the value of probable error of each of the separate observations
and an erroneous
result has
been deduced.
The
computer, apparently, has had his attention engrossed by
"
minimum
tained
for
squares
"
as the important result to be ob
whereas, in reality, the satisfying of the equations
minimum
squares produces a merely accidental coinci
dence of results in certain cases (not in all) with those
leading to the "
which
is
(2)
probable error of
final result,"
In some instances, entangled observations have
been treated as
result has
(3)
minimum
the legitimate object of search.
been
if
they were independent, and an erroneous
inferred.
In some instances, the labour of application of the
theory has been greatly and unnecessarily increased by the
use of numerical coefficients proceeding to several places
of decimals
possessing
We
all
when simple
factors
would have given results
desirable accuracy.
believe that,
by avoiding these
errors,
and by
otherwise conforming to the principles of this Treatise,
the Theory of
the Combination of Observations may,
without great labour, be
made a
putation of Physical Measures.
valuable aid in the com
APPENDIX.
PRACTICAL VERIFICATION OF THE THEORETICAL LAW
FOR THE FREQUENCY OF ERRORS.
With
the view of examining the practical accuracy of the
formula Axe~c2 .8x (see Article 24 and preceding Articles)
for the
frequency of the occurrence of Errors of Measure &c.
between the errormagnitudes x and x +
&x, I
collected the
number) of all the observations of the N.P.B.
of Polaris made at the Royal Observatory in the years 1869
to 1873, as reduced in each year to exhibit the mean N.P.D.
at the beginning of the year from every observation in the
In every separate year the difference between each of
year.
these mean N.P.D. and the annual mean of all was taken.
Prom the large number of observations in each year, and from
the perfect certainty as to the elements of reduction from one
year to another, it was evident that there would be no appreresults (636 in
ciable error in considering these " differences
from the mean
in each year as identical with the " diffei'ences from the
of
all
which would have been obtained
if all
"
mean "
the results had
been referred to one epoch of time and treated in one group.
I therefore extracted from the various groups all the " differences
from the mean," and arranged them in order of magnitude,
from the largest negative "difference" 2"*35 to the largest
positive "difference"
3"*51.
veritable errors of observation.
was215"
88,
These
may
be considered as
The sum of negative
errors
and the sum of positive errors +213"*73
(the
APPENDIX.
small discordance between
the
in
succeeding
215"88 + 213"73
^rp;
ooo
them
decimal
was
in Article 31, the
115
arising from the loss of figures
The
places).
* eK
= 0755
mean
,.,
error
or
from which, by the table
Modulus was found
to be 1" 1973,
,
and the
Probable Error 0"'57 11.
The errors were then divided into small groups, each group
extending over an errorrange of 0" 05 ; from 0" 03 to 0" 07,
#
0"*08 to 0" 12, 0"13 to 0""17, and so on, both in the positive

and
But, as the frequency of errors
in the negative direction.
was very small, one group was extended
 2"13, one from  2"12 to  l"98, one from
for the large values
from  2"38 to
 1"'72 to  l"58, one from + l"58
1"73
to + l"72, one from +
to + l"97, one from + l"98 to
+ 2"12, one from + 2"\L3to + 2"38, one from +2"39 to +3"58.
l"97 to l"73, one from
The only
result extracted
from these groups was, the number
in each group
of observations
and
this
was considered as
representing the frequency through an errorrange of 0"'05,
corresponding in formula to the magnitude of the central error
of the group as the independent variable
thus the number of
between 0""63 and 0"*67 was taken to represent the
frequency through an errorrange of 0"05, which must correspond in any mathematical formula to the independent
errors
For those
variable 0""65.
cases in
which longer groups were
employed, the actual number of observations was reduced so
as to
make
it
extending from
range of 0""25,
actual
number
number of
justly comparable with the
vations in other parts of the series of groups
2"'38
or five
to 2"13,
obser
thus in the group
which extends over an error
times the ordinary errorrange, the
of observations was divided by 5 to
make
it
comparable with the others, and the resulting quotient was
held to correspond to the error or independent variable
2"
30.
APPENDIX.
116
was no marked
It soon became evident that there
dis
cordance between the laws of distribution for positive and for
negative values
and therefore the corresponding numbers
Then, in order to remove small irregunumber was added to the second, the second
were added together.
larities,
the
first
to the third,
&c, and then the
first
sum was added to the
sum of the second
second original number
second, the second to the third, &c, the first
order being held to correspond to the
The extreme first and last were adopted
The numbex^s thus formed are evidently, on
The
the whole, eight times as large as the original numbers.
numbers thus produced were laid down in a graphical repre
and
so throughout.
without change.
sentation,
in
which the abscissa was the magnitude of the
observation, and the ordinate was
"difference," or error of
the corresponding
number
Then
for eight times the frequency.
a free curve was drawn by hand passing through the points
And this terminated the rerepresenting these numbers.
It appeared that, in the
ference to the facts of observation.
handdrawn curve, the ordinate
for error
might be taken
as 124.
For a similar exhibition of the
numbers given by the formula
11973, and where
results of theory, or of the
x e
<?
evidently =124,
it
where
= Modulus
was only necessary
to calculate the expression,
**
'
(1
frequency = 124 xe
1973)
or
log.
frequency =20934217
^^r^
x (Error)
2
.
This calculation was made for every value of Error
010, 015,
<fcc.
to 1G5,
and then
for 175, 210, 230, 250.
05,
APPENDIX.
The comparison of the observed and the
is
Observation.
of
Error
given in the followin table.
117
theoretical results
number of errors in each group ranging
through 0"05 of magnitude, multiplied by the factor 8.
Ordinate representing the
APPENDIX.
It is evident that the
119
formula represents with
all practicable
accuracy the observed Frequency of Errors, upon which
all
applications of the Theory of Probabilities are founded
the validity of every investigation in this Treatise
is
or THE
'IYERSITY
or
C. J.
the
and
thereby
established.
CAMBRIDGE: PRINTED BY
CLAY. M.A..AT TIIE UNIVERSITY PRESS.
/<
RETURN
TO^
CIRCULATION DEPARTMENT
202 Main Library
LOAN PERIOD
HOME USE
U.C.
BERKELEY LIBRARIES
0021050303
*$**jr
'
$$ti'fiy.*4