Sie sind auf Seite 1von 7

APPLICATION OF STATISTICAL CONCEPTS IN THE DETERMINATION OF

WEIGHT VARIATION IN SAMPLES SUBMITTED BY JOSHUA PAUL L. COMIA


J. P. COMIA1
1NATIONAL

INSTITUTE OF MOLECULAR BIOLOGY & BIOTECHNOLOGY, COLLEGE OF SCIENCE


UNIVERSITY OF THE PHILIPPINES, DILIMAN, QUEZON CITY 1101, PHILIPPINES
DATE SUBMITTED: 19 NOVEMBER 2013
DATE PERFORMED: 14 NOVEMBER 2013

ABSTRACT
Accuracy and precision are critical concepts to the performance of a laboratory to produce sound analytical results (Singer, 2001).
These are evaluated through the application of various statistical concepts in testing the results gained from experimentation. Provided
that, the physical equipment used in determining the weight of an individual 25-centavo coin from a set of 10 pieces was a modern
analytical balance. Additionally, the method of weighing by difference was utilized for the experiment. This involved taking a 25-centavo
piece from the watch glass, using forceps, holding the 10 pieces of coins and observing the number displayed by the balance.
Consequently, the absolute value of the number displayed was taken and recorded as the weight of the particular coin. Resetting the
balance memory to zero, each step was repeated until the last coin was taken from the watch glass and its weight was determined.

RESULTS AND DISCUSSION


This experiment required the use of a physical
equipment, the analytical balance, to determine the
weight of every individual 25-centavo coin from a set
of 10 pieces which were stacked on the concave
surface of a watch glass. This type of balance is
accurate in measuring to 0.0001g without hassle as it
balances a given sample by a magnetic force which is
closley related to mass (Slowinski et al, 2011).
Alongside the analytical balance, a watch glass
used to hold the 10 pieces of coins and forceps
used to take the coins from the watch glass were
also used.
The accuracy of the balance was ensured by pressing
the control ON TARE button. The screen was
observed until the blinking s came to a stop and the
displayed numbers remained on 0.0000g as this
shows stability (Slowinski et al, 2011). In the case that
the 0.0000g did not show, the ON TARE button was
pressed repeatedly until it did. Having stacked the
coins on top of each other and having some laid out
on the surface of the watch glass, the right sliding
window of the analytical balance was opened to serve

as an entrance to the platform wherein the sample


was to be placed on.
Since the analytical balance has three sliding
windows: the right side, the left side, and the top, the
dominant hand determined which side of the balance
to open. In this case, it was the right side.
Additionally, The top sliding window is used for vials
and other glassware which could not fit through the
left side and the right side of the window.
After the watch glass has been placed on the platform
and the window has been closed, the ON TARE
button was pressed once again. This enables the
balance to record the weight of the watch glass + the
ten 25-centavo coins and save it in its memory. After
stability has been attained, the sliding window is
opened and a 25-centavo coin was picked up.
Ensuring that the analytical balance was shut, the
screen showed a negative number after 5-10
seconds. The absolute value of this number was
recorded as the weight of that particular coin. The
negative number signfied the loss of the original
weight of the watch glass + the ten 25-centavo coins
that the balance stored in its memory. Subsequently
the balance was reset to zero by pressing the ON

TARE button and the steps utilized such as the


opening of the sliding window, taking of a coin, and
recording its weight was repeated until every 25centavo coin was picked up from the watch glass. It
was highly appropriate to ensure that the reading of
the balance was at 0.0000g as other people used it
after myself.
The various data gathered from the analytical
balance test showed the evident difference in their
weight. There is no similarity that existed in terms of
the results achieved. Provided that, the samples have
been divided into two data sets. Data Set 1 involved
the results from samples 1-6 and Data Set 2 involved
the total experiment which ranged from samples 1-10
shown in table 1 for statistical discussions and
computations.
Table 1. Measured weight of 25-centavo coin by the
use of an analytical balance
Sample No.

Qexp =

3.5955

3.5701

3.6023

3.7590

3.5691

3.5581

3.6103

3.5962

3.5696

10

3.5935

D
a
t
a

S
e
t

D
a
t
a

S
e
t
1

Before comparisons were made, the highest and


lowest values of each data set were taken into
consideration and evaluated as outliers. This was
done using the equation of Dixons Q-test (1) where
Xq is the suspected value, Xn is the value closest to
the suspected value, R is the range, and Qtab is the
tabulated Q-value at a certain confidence level. In the
experiment, the confidence level was 95%. This
meant that the Qtab for Data Set 1 is 0.625 while 0.466
for Data Set 2. If the calculation for Qexp, shown in
Appendix A, is greater than the value of Qtab, then the
suspected value is considered to be an outlier. As
outliers, these values are rejected and removed from
the calculations shown in Appendix B.

(1)

Table 2. Determination of outlier values


in two data sets with 95% confidence level
Suspected

Qtab

Qexp

Conclusion

H: 3.759

0.625

0.7800

OUTLIER

L: 3.5581

0.625

0.05475

NOT AN
OUTLIER

Qtab

Qexp

Conclusion

H: 3.759

0.466

0.7402

OUTLIER

L: 3.5581

0.466

0.05475

NOT AN
OUTLIER

Data Set
Values


Suspected
Data Set
Values

Weight, (g)

|!! !!! |

After the determination of the outlier, which turned out


to be 3.759, the computations, all of which are shown
in Appendix B. for various statistical concepts were
pursued. These used the equations for the sample
mean (1), standard deviation (2), relative standard
deviation (3), range (4), relative range (5), and the
confidence limits (6) whose values are shown in
table 3.
X-bar =

!"# !" !"" !"#$%& !" !!! !"#" !"#

s =
RSD =

!".!" !"#" !" !!! !"#" !"#


! (!! !"#$%& !"#$)!
!!!

!!!
!
!"#$%& !"#$

(1)

(2)

1000 3

R = 4
RR =

!
!"#$%& !"#$

1000 5

CL =

!"
!

factors such as the sample size, the confidence level


and the actual data given (Cox et al, 1974).

(6)

Table 3. Computed statistical values for the


evaluation of accuracy and precision at 95%
confidence level
Data
Set

X-bar

RSD

RR

CL

3.579

0.01890

5.281

0.0442

12.35

3.5563.603
3.571-

3.585

0.01831

5.107

0.0522

14.56
3.600

The statistics gathered for both data sets varied by


small decimal increments. The possible contributing
factor is the number of sample sizes because Data
Set 2 had more observable values than Data Set 1.
As shown in table 3, the mean for data sets 1 and 2
are 3.579 and 3.585, respectively. Though they vary,
the values are set by a close difference. As the
sample mean, they reflect the population mean in
terms of accuracy and estimate the true population
parameter (Skoog, 2014). Given that Data Set 2
included a larger sample size, it is more accurate than
Data Set 1. On the other hand, the standard deviation
of data sets 1 and 2 are 0.01890 and 0.01831,
respectively. These represented the spread of data
and helped in determining the precision of values to
the sample mean. It is observed that the larger the
sample size, the smaller the standard deviation for
changes in the data set tends to become closer to the
sample mean (The importance of n [sample size] in
Statistics, n.d.).
Further discussion involves the confidence limits of
data sets 1 and 2 which are 3.556 3.603 and 3.571
3.600, respectively. Given that the confidence level
is at 95%, the confidence limits were constricted to a
narrow range as the two values vary inversely
which means that as the confidence level increases,
the range of the confidence limits decreases or, in
other words, becomes smaller. It is the range that
might contain the true population parameter
(Confidence Limits, n.d.). Though it is not certain,
the credibility of the confidence limits are affected by

Lastly, a possible source of error is the effect of


sudden motion to the analytical balance. Instead of a
constant 0.0000g reading, the displayed number
changes +/- 0.0001g. Though this pertained to a small
margin of error, it is taken into account because the
goal of analytical experiments are to produce
accurate and precise results.
SUMMARY AND CONCLUSIONS
Statistical concepts and their significance to the
results were applied from the moment the analytical
balance displayed the weight of the first 25-centavo
coin. Success was achieved as proper handling of the
equipment was considered throughout the
experiment. Moreover, the environment of the
balance prevented moisture from affecting the data
which may have contributed errors in calculations.
Since the analytical balance seemed intact and in
proper calibration, future recommendations are not
highly necessary. One thought to keep in mind is to
consider all possible sources of error and proper
methodology.
REFERENCES
[1] Confidence Intervals. (n.d.). Statistical Help.
Retrieved
November
17,
2013,
from
http://www.statsdirect.com/help/default.htm#basics/co
nfidence_interval.htm
[2] Cox, D. R., & Hinkley, D. V. (1974). Confidence
Intervals. Theoretical statistics (p. 49; 209). London:
Chapman and Hall.
[3]
SHOP
ESSENTIALS
TRAINING
STATISTICS 220. (n.d.). What is the definition of
sample mean?. Retrieved November 17, 2013, from
http://www.toolingu.com/definition-800220-35808sample-mean.html
[4] Singer, D. C. (2001). Critical Laboratory
Operations. A laboratory quality handbook of best
practices (pp. 206-207). Milwaukee, WI: ASQ Quality
Press.

[5] Skoog, D. A. (2013). Statistical Treatment of


Random Errors. Fundamentals of analytical chemistry
(9th ed., [International ed., p. 99). Belmont, CA.:
Brooks/Cole, Cengage Learning.
[6] Slowinski, E. J., Masterton, W. L., & Wolsey, W. C.
(1973). Making Measurements - Laboratory
Techniques. Chemical principles in the laboratory (2d
ed., pp. 358-359). Philadelphia: Saunders.
[7] The importance of n (sample size) in Statistics.
(n.d.). The importance of n (sample size) in Statistics.
Retrieved
November
17,
2013,
from
http://www.conceptstew.co.uk/PAGES/nsamplesize.ht
ml

APPENDIX A
CALCULATION OF POSSIBLE OUTLIERS USING DIXONS Q-TEST
Data Set 1
Qtab = 0.625
Suspected Values: H: 3.759; L: 3.5581
H: 3.759
|!! !!! |

Qexp =
=
!
Qexp = 0.7800
Qexp > Qtab

|!.!"#!!.!"#$ |
!.!""#

|!.!"#$|
!.!""#

= 0.77999

CONCLUSION: OUTLIER
L: 3.5581

|!! !!! |

Qexp =
=
!
Qexp = 0.05475
Qexp < Qtab

|!.!!"#!!.!"#$ |
!.!""#

|!!.!""|
!.!""#

= 0.05475

CONCLUSION: NOT AN OUTLIER


Data Set 2
Qtab = 0.466
Suspected Values: H: 3.759; L: 3.5581
H: 3.759

|!! !!! |

Qexp =
=
!
Qexp = 0.7402
Qexp > Qtab

|!.!"#!!.!"#$ |
!.!""#

|!.!"#$|
!.!""#

= 0.74016

CONCLUSION: OUTLIER
H: 3.5581

|!! !!! |

Qexp =
=
!
Qexp = 0.05475
Qexp < Qtab

|!.!!"#!!.!"#$ |
!.!""#

CONCLUSION: NOT AN OUTLIER





|!.!""|
!.!""#

= 0.05475

APPENDIX B
CALCULATION OF STATISTICAL VALUES FOR EVALUATION OF ACCURACY AND PRECISION
Data Set 1
Degrees of freedom (n-1): 4
Sample Mean (X-bar)
X-bar =

!"# !" !"" !"#$%& !" !!! !"#" !"#


!".!" !"#" !" !!! !"#" !"#

(!.!"!!!!.!"#$!!.!"#$!!.!"#$!!.!!"#)
!

!".!"#$

=
= 3.57902
!
= 3.5790
Standard Deviation (s)
! (!! !"#$%& !"#$)!
!!!

s =
=

!!!

3.59553.579 2 + 3.57013.579 2 + 3.60233.579 2 + 3.56913.579 2 + 3.55813.579 2


51

!.!"#$ ! ! !!.!!"# ! ! !.!"## ! ! !!.!!"" ! ! !!.!"!# !


!

= 0.018902

= 0.01890
Relative Standard Deviation (RSD)
RSD =

!
!"#$%& !"#$

!.!"#$!

1000 =

!.!"#

1000 = 5.2808 = 5.281

Range (R)
R = = 3.6023 3.5581 = 0.0442
Relative Range (RR)
RR =

!
!"#$%& !"#$

1000 =

!.!""#
!.!"#

1000 = 12.3498 = 12.35

Confidence Limit (CL) at 95% Confidence Level


where t95% = 2.78
CL =

!"
!

= 3.579

Upper confidence limit = 3.579 +


Lower confidence limit = 3.579
CL = (3.556 3.603)

!.!"#"$!
!
!.!"#"$#
!

(!.!")(!.!"#$!)
!

= 3.579

!.!"#"$!

= 3.579 + 0.02350 = 3.603


= 3.579 0.02350 = 3.556

Data Set 2
Degrees of freedom (n-1): 8
Sample Mean (X-bar)
X-bar =

!"# !" !"" !"#$%& !" !!! !"#" !"#


!".!" !"#" !" !!! !"#" !"#

(!.!"!!!!.!"#$!!.!"#$!!.!"#$!!.!!"#!!.!"#$!!.!"#$!!.!"#"!!.!"#!)

!".!"#$

=
= 3.58496
!
= 3.585
Standard Deviation (s)
5
=1

s =

Xi sample mean
1

3.59553.585 2 + 3.57013.585 2 + 3.60233.585 2 + 3.56913.585 2 + 3.55813.585 2


+ 3.61033.585 2 + 3.59623.585 2 + 3.56963.585 2 +(3.59353.585)^2

91

!.!"!# ! !(!!.!"#$)! ! !.!"#$ ! ! !!.!"#$ ! ! !!.!"#$ ! ! !.!"#$ ! ! !.!""# ! ! !!.!"#$ ! !(!.!!"#)^!


!

0.00268291
=
= 0.018313
8
= 0.01831

Relative Standard Deviation (RSD)


RSD =

!
!"#$%& !"#$

1000 =

!.!"#$"
!.!"!

1000 = 5.1074 = 5.107

Range (R)
R = = 3.6103 3.5581 = 0.0522
Relative Range (RR)
RR =

!
!"#$%& !"#$

1000 =

!.!"##
!.!"!

1000 = 14.561 = 14.56

Confidence Limit (CL) at 95% Confidence Level


where t95% = 2.31
CL =

!"
!

= 3.579

Upper confidence limit = 3.579 +


Lower confidence limit = 3.579
CL = (3.571 3.600)

!.!"##$%&
!
!.!"##$%&
!

(!.!")(!.!"#$")
!

= 3.579

!.!"##$%&
!

= 3.585 + 0.01410 = 3.5991 = 3.600


= 3.585 0.01410 = 3.5709 = 3.571

Das könnte Ihnen auch gefallen