1
00:00:00,025 --> 00:00:06,970
[MUSIC]

2
00:00:06,970 --> 00:00:11,555
In this video we will illustrate how
to use regression analysis in place of

3
00:00:11,555 --> 00:00:12,740
at t-test.

4
00:00:12,740 --> 00:00:17,380
We will begin with a question, and
the question is, is there a statistically

5
00:00:17,380 --> 00:00:22,110
significant difference in teaching
evaluation scores for men and women?

6
00:00:22,110 --> 00:00:26,841
When we compute, the averages while
using the teaching evaluation data set,

7
00:00:26,841 --> 00:00:31,571
we find that the teaching evaluation
score for women is around 3.9, and for

8
00:00:31,571 --> 00:00:33,448
men, it's around 4.06.

9
00:00:33,448 --> 00:00:37,589
The question is, is this difference,
even though it's small,

10
00:00:37,589 --> 00:00:39,563
statistically significant.

11
00:00:39,563 --> 00:00:44,299
We can run at t-test using Python and
compute the statistical significance for

12
00:00:44,299 --> 00:00:45,180
the t-test.

13
00:00:45,180 --> 00:00:49,740
Here, our conclusion is that the teaching
evaluation scores difference between men

14
00:00:49,740 --> 00:00:53,010
and women, is statistically significant.

15
00:00:53,010 --> 00:00:56,635
What if we were to do the same
thing with the regression model?

16
00:00:56,635 --> 00:00:59,223
We will do the linear
regression in Python.

17
00:00:59,223 --> 00:01:01,830
We will be using the statsmodel library.

18
00:01:01,830 --> 00:01:06,360
We will create a list for the independent
variable, that is the female variable,

19
00:01:06,360 --> 00:01:09,160
which has been turned
to a binary variable,

20
00:01:09,160 --> 00:01:12,800
where 1 equals female and 0 is male.

21
00:01:12,800 --> 00:01:15,951
We will also create a list for
the dependent variable,

22
00:01:15,951 --> 00:01:17,674
teaching evaluation score.

23
00:01:17,674 --> 00:01:22,212
We will manually add the constant
beta zero, then we will fit and

24
00:01:22,212 --> 00:01:26,010
make predictions, and
print out the model summary.

25
00:01:26,010 --> 00:01:30,624
The model summary will print
out a table like this.

26
00:01:31,800 --> 00:01:36,324
But we are only interested in this
part of this table for the t-test,

27
00:01:36,324 --> 00:01:40,700
it prints out the coefficient error,
t statistics and pin value.

28
00:01:40,700 --> 00:01:45,666
We can see the t statistics for
the female variable is negative 3.25,

29
00:01:45,666 --> 00:01:48,242
and the P value is less than 0.05.

30
00:01:48,242 --> 00:01:53,220
That means that there is a statistical
difference in mean values for male and

31
00:01:53,220 --> 00:01:54,710
female instructors.

32
00:01:54,710 --> 00:01:59,415
The coefficient means that you are most
likely to lose about 0.17 marks for

33
00:01:59,415 --> 00:02:00,421
being a female.

34
00:02:00,421 --> 00:02:03,928
We can see that the results from
using a regression model, and

35
00:02:03,928 --> 00:02:06,842
the conclusion is identical,
if we run a t-test.

36
00:02:06,842 --> 00:02:07,342
[MUSIC]