DASHBOARD LEARN MENU

Learn VEE Mathematical Stats 1 1.2 1.2.3 Student's t-Distribution

Student's t-Distribution

Let's resume where we ended with statistic X̄ having a normal sampling distribution. Recall that standardizing X̄
gives the standard normal random variable.

X̄ − E[X̄] X̄ − μ X̄ − μ
Z= −−−−−−− = −−−− =
√V ar[X̄] √σ 2 /n σ/√− n

Now imagine that the value of σ 2 is not known. In its place, substitute in the sample variance, S 2 . This
produces the statistic

X̄ − μ
S/√− n

X̄−μ
If the X 's follow a normal distribution, then the sampling distribution of is a Student's t -distribution with
S /√n
n − 1 degrees of freedom.

Defining characteristics of this statistic include:

• Bypasses the need to know σ 2

• Similar to a standard normal random variable

The t -distribution is a continuous distribution with one parameter called degrees of freedom. It is very similar to a
standard normal distribution; its density function is also symmetric about 0. It is primarily different from a standard
normal by having a larger kurtosis (i.e. higher chance for outliers). In fact, as the degrees of freedom approaches
∞ , a t -distribution converges to a standard normal distribution.

By definition, for a standard normal random variable Z independent of a χ2 random variable Y with ν degrees of
freedom, the expression

Z
−−−−
√Y / ν

has a t -distribution with ν degrees of freedom.

X̄−μ (n−1)S 2
For a random sample drawn from a normal distribution, it can be proven that is independent of ,
σ /√n σ2
thus showing that

−−−−−−−−−
√
X̄ − μ (n − 1)S 2 X̄ − μ
− ÷ =
σ/√n (n − 1)σ 2
S/√− n

is t -distributed with n − 1 degrees of freedom.

As was done for χ2 , we will provide a small table of t percentiles in each problem when necessary.

Let's try several examples.

EXAMPLE 1.2.4

A teacher finds test scores to be normally distributed with mean 82. She randomly selects 4 students from the
class and checks their test scores.

Let tp, ν be the 100p th percentile of a t random variable with ν degrees of freedom. The following table lists
values of tp, ν for specific combinations of p and ν :

p = 0.8627 p = 0.8965 p = 0.9303

What is the probability that the sample mean of the 4 scores minus 82 is greater than -0.802 times the sample
standard deviation of the 4 scores?

SOLUTION

Since the sample size is 4, the appropriate t -distribution has 4 − 1 = 3 degrees of freedom.
 Pr (X̄ − 82 > −0.802S)
= Pr (X̄ − μ > −0.802S)

= Pr ( – )
X̄ − μ −0.802S
− >
 S/√n S/√4 

(
Like the standard normal distribution, the t -distribution=is Pr X̄ −about
symmetric μ 0. Thus,
> −1.604)
S/√− n
Pr ( > −1.604) = Pr (
X̄ − μ X̄ − μ
− < 1.604)
S/√n S/√− n
= 0.8965

EXAMPLE 1.2.5

Heavenly Airlines is testing the efficiency of its new type of airplane. A manager believes that a flight time without
refueling is normally distributed with mean μ . A researcher gathers a random sample of 20 airplanes to test.

The probability that the sample mean is within 0.468 sample standard deviations from μ is 0.95.

Calculate the probability that the sample mean minus μ is less than 0.468 sample standard deviations.

SOLUTION

The goal is to calculate

Pr (X̄ − μ < 0.468S) = Pr ( −− )

= Pr (
X̄ − μ
< 2.093)
S/√− n

We are told that


⇒ Pr (−0.468S < X̄ − μ < 0.468S) = 0.95 

Pr (−2.093 <
X̄ − μ
< 2.093) = 0.95
Translating into a graph, S/√− n

Hence, the probability in the tails sum to 5%. Since 5% must be equally divided between the two tails,
These graphs show that the answer is

Pr (
X̄ − μ
< 2.093) = 0.025 + 0.95 = 0.975
S/√− n

COACH'S REMARKS

We can generalize this result in math notation:

For a continuous random variable Y that is symmetric about 0,

1−k
Pr (−a < Y < a) = k ⇒ Pr (Y < −a) =
2
1+k
⇒ Pr (Y < a) =
2

This will be useful for the material on confidence intervals.

Discussions

Ask a question

Nur Alia Kamaluddin

SUMMARY:

MESSAGE:

Type your question...