0
$\begingroup$

I wonder if you could give me a hint on how to find out whether there is a significant difference in the income gap between genders and their education level. The picture below shows education levels (edu_level_old), means for women and men income (in Russian currency), and a gap in income measured by %.

enter image description here

I had no problem with finding out whether, for example, women with higher education earn less than men with higher education (done this by ttest). BUT what I would like to discover is whether this gap in earnings between different education levels (for example, a gap between those with higher education and secondary education) is statistically different. Is there a way I could do that? I have a feeling that it is rather simple and obvious for some but have no idea how to accomplish that. Your help is much appreciated!

$\endgroup$

1 Answer 1

0
$\begingroup$

There are three main types of t-test:

  • An Independent Samples t-test (IS t-test) compares the means for two groups.
  • A Paired sample t-test (PS t-test) compares means from the same group at different times, or occasions
  • A One sample t-test (OS t-test) tests the mean of a single group against a known mean.

What type of the t-test you have used in your previous work? Was it R or Python? Because I see a problem, you have means but you don't have N. You cannot calculate the t-test without N.

Example in Python

N=100
a = np.random.randn(N) + 1
b = np.random.randn(N)

var_a = a.var(ddof=1)
var_b = b.var(ddof=1)

#std deviation
s = np.sqrt((var_a + var_b)/2)
s

## Calculate the t-statistics
t = (a.mean() - b.mean())/(s*np.sqrt(2/N))

## Compare with the critical t-value
#Degrees of freedom
df = 2*N - 2

#p-value after comparison with the t 
p = 1 - stats.t.cdf(t,df=df)

It looks to me that the gap in income is useless information. Right?

I would like to discover is whether this gap in earnings between different education levels (for example, a gap between those with higher education and secondary education) is statistically different.

I think your problem is the gaps are not means so you cannot run t-tests on them. So the issue is that t-test isn't appropriate for percentages, nothing you can do about it.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.