Dr. Md.

Atiqul Islam
PhD (UK), MSc (UK)
Professor
Department of Economics

University of Rajshahi

Course: ECO 314 (Statistics for Economics III)

05 November 2024
The 𝑡-Test and F-Test

Part I
The t-Test:
• For small samples (at most 30), the sample variance can be quite different
from the population variance. Therefore, it will involve considerable error if
we use the z test.
• Moreover, when sample is small then
𝑥−𝜇
𝑠 ∼ 𝑡𝑛−1
𝑛
• Thus when the sample is small, we need to use the t-Test for inference about
the mean of the population.
• Thus to test the hypothesis 𝐻𝑜 : 𝜇 = 𝜇0 against 𝐻1 : 𝜇 ≠ 𝜇0 we need to use t
statistics as
𝑥−𝜇
𝑡= 𝑠
𝑛
• At 𝛼 significance level, reject 𝐻𝑜 if calculated 𝑡 > 𝑡𝛼/2 , 𝑛−1 .
The confidence interval:
• At 𝛼 significance level, the confidence interval for 𝜇 is computed as

𝑠 𝑠
𝑥 − 𝑡𝛼/2, 𝑛−1 < 𝜇 < 𝑥 + 𝑡𝛼/2, 𝑛−1
𝑛 𝑛
The t-distribution
Let 𝑥1 , 𝑥2 , … . . 𝑥𝑛 be a random sample of size n from a normal population
with mean 𝜇 and 𝜎 2 then Student’s t is defined by the statistics as
𝑥−𝜇
𝑠
𝑛
1 1
Where 𝑥 = ∑𝑥𝑖 is the sample mean and 𝑠 2
= ∑ 𝑥𝑖 − 𝑥 2
is an
𝑛 𝑛−1
unbiased estimate of the population variance 𝜎 2 , and it follows Student’s t-
distribution with 𝜈 = 𝑛 − 1 degree of freedom with probability density
function as
1
𝑓 𝑡 = 𝜈+1 , −∞ < 𝑡 < ∞
1 𝜈 𝑡2 2
𝜈𝛽 , 1+
2 2 𝜈
The t-distribution…continued
This t-distribution is used to estimate population parameters when the
sample size is small and/or when the population variance is unknown.

Properties of t-distribution
• The mean of the t distribution is equal to 0.

𝜈
• The variance is equal to where v is the degree of freedom and v>2
𝜈−2

• The variance is always greater than 1. With infinite degrees of freedom,

the t-distribution is the same as the standard normal distribution.
Example (10): A tyre manufacturer claims that the average life of their tyre is
greater than 25,000 km. A random sample of 15 shows a mean and standard
deviation as 27000 and 5000 km respectively. Check and comment about
manufacturer’s claim.
Ans. Given 𝜇 = 25000 , 𝑥 = 27000 , 𝑠 = 5000 and 𝑛 = 15
Here, 𝐻0 : 𝜇 = 25000 and 𝐻1 : 𝜇 > 25000
As the sample size is small (less than 30), to comment on the population mean
we need to use the t test. Therefore,
𝑥 − 𝜇 27000 − 25000 2000 15
𝑡= = = = 1.549
𝑠/ 𝑛 5000/ 15 5000
Here, the table value of 𝑡0.05, 14 = 1.761
As, calculated value of t < table value of t , so null hypothesis is accepted.
Therefore, the tyre is not as good as manufacturer’s claim. (Ans.)
Test of hypothesis about the difference between two means:
• For small sample, the difference between the means of two normally
distributed population the t – Test is used.
(a) When variances are equal, i.e. 𝜎12 = 𝜎22
(b) When variances are not equal, i.e. 𝜎12 ≠ 𝜎22
(a) When variances are equal, i.e. 𝜎12 = 𝜎22
Here, 𝐻0 : 𝜇1 = 𝜇2 (i.e. means are equal or there is no significance difference)
𝑥1 −𝑥2 𝑥1 −𝑥2 𝑛1 𝑛2
The test statistics is 𝑡 = = with (𝑛1 +𝑛2 − 2) d.f.
𝑠
1
+
1 𝑠 𝑛1 +𝑛2
𝑛1 𝑛2

𝑛1 −1 𝑆12 + 𝑛2 −1 𝑆22 ∑ 𝑥1𝑖 −𝑥1 2 ∑ 𝑥2𝑖 −𝑥2 2

Where 𝑠 = , 𝑠12 = , 𝑠22 =
𝑛1 +𝑛2 −2 𝑛1 −1 𝑛2 −1
Test of hypothesis about the difference between two means:

(b) When variances are unequal, i.e. 𝜎12 ≠ 𝜎22

Here, 𝐻0 : 𝜇1 = 𝜇2 (i.e. means are equal or there is no significance difference)

𝑆2 𝑆 2 2
1+ 2
𝑥1 −𝑥2 𝑛 1 𝑛2
The test statistics is 𝑡 = with 𝑑. 𝑓. = 2 2 2 2
𝑆2 2 𝑆1 𝑆2
1 + 𝑆2 𝑛1 𝑛2
𝑛1 𝑛2 +
𝑛1 −1 𝑛2 −1
Example (13): The mean life of a sample of 10 electric light bulb 1,456 hours
with variance of 423 hours. A second sample of 17 bulb chosen from second
batch shows a mean life of 1280 hours with variance of 398 hours. Is there a
significance difference between the mean of two batches.
Ans.
Here 𝑥1 = 1456, 𝑠12 = 423, 𝑛1 = 10, 𝑥2 = 1280, 𝑠22 = 398, 𝑛2 = 17,
As there are two batches, we may assume population variance are unequal.

Here, 𝐻0 : 𝜇1 = 𝜇2 and 𝐻1 : 𝜇1 ≠ 𝜇2

𝑥1 −𝑥2 1456−1280 176 176

The test statistics is 𝑡 = = = = = 21.71
2 423 398 42.3+23.41 8.106
𝑠2 𝑠
1+ 2 +
10 17
𝑛1 𝑛2
Example (13): The mean life of a sample of 10 electric light bulb 1,456 hours
with variance of 423 hours. A second sample of 17 bulb chosen from second
batch shows a mean life of 1280 hours with variance of 398 hours. Is there a
significance difference between the mean of two batches.
Ans.

Here 𝑠12 = 423, 𝑛1 = 10, 𝑠22 = 398, 𝑛2 = 17,

𝑆2 𝑆 2 2
1+ 2 423 398 2
𝑛 1 𝑛2 +
10 17
Here, 𝑑. 𝑓. = 2 2 2 2 = 423 2 398 2
𝑆1 𝑆2
10 17
𝑛1 𝑛2
10−1
+ 17−1
+
𝑛1 −1 𝑛2 −1

42.3+23.41 2 4317.8041
= = = 18.52 ≡ 19
198.81+34.25 233.06
Example (13): The mean life of a sample of 10 electric light bulb 1,456 hours
with variance of 423 hours. A second sample of 17 bulb chosen from second
batch shows a mean life of 1280 hours with variance of 398 hours. Is there a
significance difference between the mean of two batches.
Ans. Here
The test statistics is 𝑡 = 21.71

The table value of 𝑡0.025, 19 = 2.093

As the calculated value is greater than the table value so we reject the null
hypothesis.

Thus, difference between the mean of the two batches are significance. (Ans.)