1
$\begingroup$

I have two main independent variables - gender (0/1) and extraversion (continuous) in a logit regression. In the full sample, the gender X extraversion interaction shows no statistical significance. However, when I split the sample into male vs female subgroups, extraversion is statistically significant (p<0.05) among women, but not among men (p=0.09). The coefficient of extraversion in both subgroups are positive. What does this mean? Can I conclude that extraversion significantly predicts women's outcomes, but not for men? Thank you!

$\endgroup$

1 Answer 1

2
$\begingroup$

Can I conclude that extraversion significantly predicts women's outcomes, but not for men?

Yes and no. Yes it is significant in one subgroup. But, is the association significantly different between men and women? No - the interaction is not significant. What is going on? You didn't tell us the main effect of extraversion, but probably it is also significant? The fact that you don't have a significant interaction means that this association doesn't differ between men and women. Probably it is just non-significant in men due to the loss of power, because the sample is smaller.

$\endgroup$
3
  • $\begingroup$ Thanks! Yes, the main effect for Extraversion is also significant. The sample of men is actually 2x as large as that of women. So I guess this may indicate the extraversion is just not as important for Dependent Variable for men as for women? $\endgroup$ Apr 21 at 3:59
  • $\begingroup$ Possibly. But again, if you're using p-values, then you can't say much other than 'not significant'. It could still just be low power + randomness. $\endgroup$
    – David B
    Apr 21 at 12:18
  • $\begingroup$ +1 I would add that the analysis approach is suspect (this may not be done intentionally): You fit a model with an interaction presumably because you have an interest in investigating differences between men and women. The data supports no such conclusion. So then you try other ways to demonstrate a difference by subsetting the data. $\endgroup$
    – dipetkov
    Apr 21 at 21:08

Your Answer

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

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