2
$\begingroup$

I'm having difficulty interpreting the interaction term is a linear regression. I'm using two dichotomous variables, for instance 1 or 0 for gender and 1 or 0 for type of class - active or passive, coded as 1 or 0. The stata results reported the result for when the two variables are both 1, which is male 1 and active class 1. However, the coefficient estimate are 0 for all other cases (male 1 and passive class 0; female 0 active class 1; female 0 passive class 0).

I think the results are telling me that one variable - type of class only explains the variation explained by gender only in the case when the gender is male. Am I wrong? If I am right, is there value to such analysis? I currently think there is, but I am also not sure why stata reports 0 for the rest of the cases.

$\endgroup$
2
$\begingroup$

A concrete example is easier to follow. Let us say the dependent variable is average score in math. When an interaction term exists between two binary variables, its coefficient means the the mean added (math score) for a student who is both male and has an active class, above and beyond just being male and just having an active class.

So, if I understand your description, in which all coefficients except the interaction are zero, it means that being male has no significant difference (in math score) than being female, and that having an active class has no significant difference (on math score) than having a passive class, but that being male AND having an active class does have a difference (I guess significant, but it wasn't clear to me) on the dependent variable (math score).

Afterthought - if you can post the regression output, this can be much more specific an answer.

$\endgroup$
  • $\begingroup$ Yes, I have drawn a map plotting the predicted outcomes of the regression plot and it does seem the only thing matters was being male and having an active class. The coefficient was also statistically significant. $\endgroup$ – user2298759 Jun 6 '18 at 16:08
  • $\begingroup$ Although, I am trying to grapple with the idea of reporting the results - I'm thinking that this implies that there is something going on within the male group in regards to participating in active or passive class. $\endgroup$ – user2298759 Jun 6 '18 at 16:09
  • $\begingroup$ There is something going on. Statistics are great, but interpretation is also important, and this can be done based on previous literature, or your own ideas - if they can be supported by the data, and alternative explanations ruled out. $\endgroup$ – Yuval Spiegler Jun 6 '18 at 16:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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