question about t-test function

[Tundra.L]

Up

0

Down

When I did t-test, the parameters affect lot, the results are different. I am not sure which one is better:

t,p_value=stats.ttest_ind(Loan[Loan['Bad']==0]['mtg_due'],
                          Loan[Loan['Bad']==1]['mtg_due'],
                           nan_policy='omit')
print("ttest_ind:        t = %g  p_value = %g" % (t, p_value))
#t = 4.45713  p_value = 8.40104e-06

t,p_value=stats.ttest_ind(Loan[Loan['Bad']==0]['mtg_due'].isnull()==False,
                          Loan[Loan['Bad']==1]['mtg_due'].isnull()==False)                                        
print("ttest_ind:        t = %g  p_value = %g" % (t, p_value))
#t = 1.17903  p_value = 0.238412

t,p_value=stats.ttest_ind(Loan[Loan['Bad']==0]['mtg_due'],
                          Loan[Loan['Bad']==1]['mtg_due'],
                          equal_var=False, nan_policy='omit')                  
print("ttest_ind:        t = %g  p_value = %g" % (t, p_value))
#t = 4.19483  p_value = 2.82717e-05

the results are so different, it causes if the H0 is rejected or not.

[Justin]

Up

0

Down

Generally, you should remove the null values and use non-equal variances.

[Tundra.L]

Up

0

Down

thank you.

non-equal variance means we should use:

equal_var=False

correct?

[Justin]

Up

0

Down

Yes, correct!

[Tundra.L]

Up

0

Down

thank you