Derivation of the Normal Distribution:

Lecture VI
I. Derivation of the Normal Distribution Function
A. The order of proof of the normal distribution function is to start with the standard
normal:
f x  
1  x2 2
e
2

1. First, we need to demonstrate that the distribution function does integrate

to one over the entire sample space, which is  to  . This is typically
accomplished by proving the constant.
2. Let us start by assuming that
2
 y
I  e 2
dy


Squaring this expression yields

2
 y  x
2

I2   e 2
dy  e 2
dx
 
y2  x2
  
  e 2
dy dx
 

The trick to this integration is changing the variables into a polar form.
3. Polar Integration: The notion of polar integration is basically one of a
change in variables. Specifically, some integrals may be ill-posed in the
traditional Cartesian plane, but easily solved in a polar space.
a. By polar space, any point  x, y  can be written in a trigonometric
form:
r  x2  y2
  tan 1 x y 
y  r cos  
x  r sin  
i. As an example, take f  x   15  x  1 x 2 . Some of the
2
results for this function are:
AEB 6933 Mathematical Statistics for Food and Resource Economics Lecture VI
Professor Charles B. Moss Fall 2007

Table 1. Polar Transformation

x f  x r 
-5.00 -2.5000 5.5902 0.4636
-4.00 3.0000 5.0000 -0.6435
-3.00 7.5000 8.0777 -1.1903
-2.00 11.0000 11.1803 -1.3909
-1.00 13.5000 13.5370 -1.4969
-0.50 14.3750 14.3837 -1.5360
-0.25 14.7188 14.7209 -1.5538
0.00 15.0000 15.0000
0.25 15.2188 15.2209 1.5544
0.50 15.3750 15.3831 1.5383
1.00 15.5000 15.5322 1.5064
2.00 15.0000 15.1327 1.4382
3.00 13.5000 13.8293 1.3521
4.00 11.0000 11.7047 1.2220
5.00 7.5000 9.0139 0.9828

ii. Adding a little more definition and plotting in Polar space

yields
r
2.5

2.0

1.5

1.0

0.5

theta
8 10 12 14

b. The workhorse in this proof is Greene’s theorem. We know from

univariate calculus that
b d f t 
a d t dt  f b  f a 
In multivariate space, this becomes
 N x  M y dx dy   Mdx   Ndy
S  

2
AEB 6933 Mathematical Statistics for Food and Resource Economics Lecture VI
Professor Charles B. Moss Fall 2007
To see this note that
  N
S x  M y dx dy   N x dx dy   M y dx dy
S S

  N dy   M dx
 

c. Picture, the general form

f  x   exp  x
2
 2 
Mapped into polar space. The original graph of the function is

1.0

0.8

0.6

0.4

0.2

4 2 2 4

The Polar form of this expression can be depicted as

3.0

2.5

2.0

1.5

1.0

0.5

theta
1.5 2.0 2.5 3.0

3
AEB 6933 Mathematical Statistics for Food and Resource Economics Lecture VI
Professor Charles B. Moss Fall 2007

Thus, letting
y

M  x, y     f x, t dt , N x, y   0

 x

f dx dy   N x  M y dx dy    M dx
4


S S i 1  i

d
f dx dy   f r dr d     f r cos  , r sin  r dr  d
b

S S c
 a 
4. By substituting y  r cos   and x  r sin   we derive
y 2  x 2  r 2 cos 2    r 2 sin 2  

 r 2 cos 2    sin 2   
r 2

Defining the derivatives of the change in variables

dy dx  r dr d
Folding these two results together, we get
2  r
2
2
I2    re 2
dr d   d  2
0 0 0
Taking the square root of each side yields

I  2

Thus, we know that

2
 1 y
 
2
e 2
dy  1

B. The expression above is the expression for the standard normal. A more general
form of the normal distribution function can be derived by defining a
transformation function.
1. Defining
y  a  bx
ya
x
b

2. By the change in variable technique, we have

  y  a 2 
  
f x  
1  2 b 2  1
e
2 b

4
AEB 6933 Mathematical Statistics for Food and Resource Economics Lecture VI
Professor Charles B. Moss Fall 2007

II. Expected Values

A. Definition 2.2.1. The expected value or mean of a random variable g  X  noted
by E  g  x  is
 
 g x  f x dx if x is continuous
Eg  X   
 xX gx  f x  if x is discrete


B. The most general form used in this definition allows for taking the expectation of
the function g  X  . Strictly speaking, the mean of the distribution is found where
g  X   X , or

Ex   x f x dx




For example, we take the exponential distribution:

E x   
 1 x
x e  dx
0 
 x 
 x
  xe   e 
dx
 0
0

 x 
  e  
 0

C. Theorem 2.2.1. Let X be a random variable and let a , b , and c be constants.

Then for any functions g1  X  and g 2  X  whose expectations exist:
1. E a g1  X   b g2  X   c   a E  g1  X   b E  g 2  X   c .
2. If g1  X   0 for all X , then E  g1  X   0 .
3. If g1  X   g2  X  for all X , then E  g1  X   E  g2  X  .
4. If a  g1  X   b for all X , then a  E  g1  X   b .