Econ 3334

Review of Probability Theory-

Part B

1
 4) Distribution Econ 3334

 Normal Distribution X  N,  2 

 Pr  1. 96  X    1. 96  0. 95

 4) Distribution Econ 3334

 In general, we standardize the Normal R.V. by subtracting 

and dividing by .

Prc  X  d

c X d c d

 Pr       Pr   Z  

where Z  N0,1
 4) Distribution Econ 3334
 Example: Income in the US is the normally distributed with
mu=50,000 and sigma=20,000. What is the probability a
person in the US has income over 75,000.
X  50, 000 75, 000  50, 000
PrX  75, 000  Pr 
20, 000 20, 000
 PrZ  1. 25  10. 56%

 What is the probability that a person at random has income

between 35,000 and 65,000?
Pr35, 000  X  65, 000
35, 000  50, 000 X  50, 000 65, 000  50, 000
 Pr  
20, 000 20, 000 20, 000
 Pr0. 75  Z  0. 75  1  PrZ  0. 75  PrZ  0. 75
 1  2  PrZ  0. 75  1  2  0. 2266  0. 5468  54. 68%
 4) Distribution Econ 3334

 Bi-variate Normal: this describes the probability distribution

of X and Y are jointly as Normal

 The linear combination aX+bY is distributed as:

N aX  bY, a 2   2X  b 2   2Y  2ab   XY

 If X and Y are independent, then aX+bY is distributed as:

N a X  b Y, a 2   2X  b 2   2Y
 4) Distribution Econ 3334
 Chi-squared distribution  2
 The distribution of the sum of k squared independent standard
normal RV is called  2k distribution , and k is called the degree
of freedom.
 Suppose that we have 4 (i.e., k=4) standard normal variable that
are independent: Z 1 , Z 2 , Z 3 , Z 4. Let W  Z 21  Z 22  Z 23  Z 24 . Then the
distribution of W is  24
 PrW  9. 49  0. 05 , see table 3 on page 758
 PDF:
 4) Distribution Econ 3334

 Student t distribution
 Let Z be a standard normal RV and Wm a Chi-squared
RV with degree of freedom m. Suppose Z and Wm are
independent.
 Then Rm  Z  t distribution with m degree of freedom
Wm
m


 4) Distribution Econ 3334
 F distribution
 Let W be a chi-squared random variable with m degree of freedom.
 Let V be a chi-squared random variable with n degree of freedom.
 W/m
 F m ,n (F distribution with m and n degree)
V/n

As n  , W/m  F m ,

5) Sampling Distribution
Econ 3334

A sample

Sampling is the process of taking a smaller

group of subjects from a larger population.
5) Sampling Distribution
Econ 3334

..
.. .
. .
.. . . .
. . Sampling
.
.. .. .
. . . (X1 ,X2 , ... ,Xn )
[A simple random sample]
. .
..
POPULATION
Statistical Inference
5) Sampling Distribution
Econ 3334

The random sample or the data { X , X , X ..., X }

1 2 3 n

are random variables.

 We don’t know the exact outcome beforehand.

The statistic calculated from a randomly chosen

sample is an example of a random variable.
A statistic from different random samples will
take different values.

11
 5) Sampling Distribution Econ 3334
 Populations Have Parameters, Samples Have Estimators

Population Sample
Parameters Estimators
e.g. Pop. mean Sample mean
Pop. Variance Sample variance
 5) Sampling Distribution Econ 3334
 Suppose
Yi  N,  2 , i  1, . . . , n
Y
 What is the distribution of Y  n i ?
 Here we focus on the case where a sample is draw at
random (random sampling) from a population
 For a sample size n=20, draw at random Y1 , Y2 , . . . , Y20 each Yi
is independent from the other since random sampling
 If they are draw from the same underling distribution (e.g.,
Normal), then they are identically distributed
 iid: independently and identically distributed.
iid
Y1 , Y2 , Y3 , . . . , Yn  N,  2 
 5) Sampling Distribution Econ 3334
 Average Y   Yi
n
 For different random sample drawn from the population
implies that Y is a random variable
 Y changes from sample to sample
 So we are interested in that variability called the sampling
distribution
 Suppose Y1 , Y2 , Y3 , . . . , Yn iid N,  2 
 Then
 Y  EY  E n  Yi  n EY1  Y2 . . . Yn 
 1 1

 1
n EY1   EY2  . . . EYn   1    . . .   
n

 2Y  varY   var 1n  Yi  12 varY1  Y2 . . . Yn 

n
 12 varY1   varY2  . . . varYn   12  2   2 . . .  2   n
2

n n
 5) Sampling Distribution Econ 3334
 Example: Incomes normally distributed mu=50,000 and
sigma=20,000. 100 people sampled at random from the
distribution: Y  N 50, 000, 20,0002
100
 What is the probability the average income is greater than
55,000?  Y  50, 000 55, 000  50, 000
PrY  55, 000   Pr 20,000
 20,000
100 100

 PrZ  2. 5   0. 0062  0. 62%

 What is the probability the average income is between 45,000

and 52,000.
45, 000  50, 000 Y  50, 000 52, 000  50, 000
Pr45, 000  Y  52, 000   Pr 20,000
 20,000
 20,000
100 100 100

 Pr2. 5  Z  1   1  PrZ  2. 5   PrZ  1   0. 8351

6) Large sample distribution Econ 3334
 We know that Yi  N,  2 , i  1, . . . , n , then Y  N , n2 .
 But if Yi  some complicated distribution, i  1, . . . , n
then Y  ? Potentially complicated!
 But
Y  
As n  ,  N0, 1 
/ n
 This is called central limit theorem (CLT).
 6) Large sample distribution Econ 3334
 Law of large number
 Y1, Y2, Y3, . . . , Yn drawn iid from distribution with mean mu.
 As n goes to infinite, Y gets closer and closer to mu.
 Convergence in probability or “consistency”
For any c  0, Pr  c  Y    c   1, as n  
p
 Stated as Y   . This is saying

PrY     1, as n  
Pr|Y  |  c   1, as n  
Pr|Y  |  c   0, as n  
 6) Large sample distribution Econ 3334
 Law of large number
 Example, Y1, Y2, Y3, . . . , Yn iid
 N  12,  2  25

n  2, Y  7. 80
n  10, Y  11. 67 As n  , Y  
n  100, Y  11. 84
n  1000, Y  11. 97

 An interactive example:
 http://digitalfirst.bfwpub.com/stats_applet/generic_stats_applet_11_largenums.htmll
 6) Large sample distribution Econ 3334
 Central limit theorem
 Y1, Y2, Y3, . . . , Yn : iid, then
Y   A
As n  ,  N0, 1  or
/ n
 A 
As n  , Y  N u, n
2

 Very powerful results. It says that even if we don’t know

the distribution of Y1, Y2, Y3, . . . , Yn , we know Y will be
normally distributed if n is sufficiently large.
 The bigger n becomes, the closer Y ‘s distribution become
normal.
 An interactive example:
http://www.mathsisfun.com/data/quincunx.html