CHAPTER 5 Distributions of Functions of Random Variables

Section 5.5 Random Samples from a Normal Distribution; t distribution

A statistic
 is a function of the random variables in a random sample, i.e., f ( X 1 , X 2 , , X n ) .
 is itself a random variable.
 has a distribution which is usually called a sampling distribution (i.e., a distribution that arises as a
result of sampling from the population or distribution).
Examples of a statistic: sample mean X , sample variance S 2 , median

Result
If X 1 , X 2, , X n are a random sample of size n from any population or distribution whose mean is  and
whose variance is  2 , then

 
E X 
Var X    2 n

Proof: (Omitted)

Result

If X 1 , X 2, , X n are a random sample of size n from a normal distribution N  ,  2  , then the sampling
distribution of the sample mean
n

X
1
X  i
n i 1


is also normal, but with mean  and variance  2 n , i.e., N  ,  2 n . 
Proof: (Omitted)

Implication:
X 
The variable Z  has a normal distribution with mean 0 and variance 1, i.e., the standard normal
 n
distribution N 0,1 .

27
 Sampling Distribution of Xbar for n = 16
Sampling Distribution of Xbar for n = 4
Sampling Distribution of Xbar for n = 2
Normal Population (Distribution of X)
f(x)

2 3 4 5 6 7 8
x

Plot of a normal distribution with   5 and variance  2  1 , and the sampling distributions of X with sample sizes
n  2, 4, 16.

Example:

Suppose the tensile strength of a block of one type of plastic material of a standard size has approximately a
normal distribution with mean 35 psi and standard deviation 4 psi. A random sample of 50 blocks of such a
material is taken. What is the probability that the average tensile strength of the sample is between 34 and
36 psi?

28
The t Distribution

Typically the true variance  2 of a population or distribution is not known, and is estimated by the sample
variance, which is defined as

 X 
n
1 2
S2  i X ,
n 1 i 1
where X 1 , X 2, , X n are a random sample from the unknown population or distribution.

Alternate forms of the sample variance are

2
n  n 
X
i 1
i
2

 
 i 1
Xi 


n
1. S2 
n 1
n

X
2
i
2
 nX
i 1
2. S2 
n 1

The sample standard deviation is given by

 X 
n
1 2
S i X .
n 1 i 1


Suppose the random sample X 1 , X 2, , X n is taken from a normal distribution N  ,  2 . Then the
statistic
X 
S n

 has a sampling distribution known as Student’s t distribution, or t distribution for short, with
n  1 degrees of freedom.

t Distribution
1. The p.d.f. is given by
 1
    1 / 2
 2  1  t 
2
f (t ; )  1
v
     
 2
where   t   ,  is known as the degrees of freedom, and  is the gamma function.

29
2. The general shape of the p.d.f. is nearly identical to the standard normal distribution, except that the t
distribution has heavier tails. It is symmetric about zero, which is also its mean.
3. As its degrees of freedom    , the t distribution approaches the standard normal distribution.

Example:
Using a t table, check that for a random variable T with t distribution with  degrees of freedom
 PT4  2.132  0.05
 P(2.086  T20  2.086)  .95
 P2.131  T15  2.947  0.02


Note: To find probabilities with numbers not on the t table, e.g. PT4  2.5 , a calculator with t distribution
or a statistical software package will have to be used.

Section 5.6 The Central Limit Theorem

In the previous section, we have assumed that, for the most part, that the random sample X 1 , X 2, , X n
 
comes from N  ,  2 . Then the statistic
X 
Z
 n
has a standard normal distribution N 0,1 .

It turns out that the random sample X 1 , X 2, , X n can be taken from any population, and as long as n is
X 
large enough, Z  has approximately the standard normal distribution.
 n

30
Central Limit Theorem

Let X 1 , X 2, , X n be independent random variables with a common distribution with mean  and
variance  2   . Then the distribution of
X 
Z
n 
converges to the standard normal distribution as n   .

Remarks:
1. A more useful way of stating the central limit theorem is: Let X 1 , X 2, , X n be a random sample
from any population with mean  and variance  2   . Then when the sample size n is large, the
distribution of X is approximately normal with mean  and variance  2 n , i.e.,
 
X is approximately distributed as N  ,  2 / n .

2. Another useful form of the Central Limit Theorem is:

 
n

X
i 1
i is approximately distributed as N n , n 2 ,

under the same conditions.

3. How large, in most applications, should the sample size n be for the central limit theorem to take
effect, i.e., for X to be approximately normally distributed?
Answer: A general rule of thumb is a sample size n of at least 30.

Example:
The fracture strength of a brittle material has an unknown distribution with mean 14 (in lb/in2) and std.
deviation 2. What is the probability that the average fracture strength of 100 randomly selected pieces of
this material exceeds 14.5?

31
Example:
A biologist who will spend 60 days in a remote part of the world on a research expedition plans to bring
enough supply of a certain type of battery for one of his instruments. He estimates that he will use
approximately 6 hours of a battery’s life every day. Suppose the lifetime of a battery has a distribution with
mean 8 hours and standard deviation 1 hour. What is the probability that a supply of 46 batteries will run
out before the expedition is over? How about 47 batteries?

32