Significant interaction term but no increase in r2

[Johan Anderson]

Hey everyone,
I was just wondering how to interpret my interaction terms in my regression analysis. They are theoretically motivated and significant at the p<0.01 level. However, they do not improve the fitness of the model (r2). Is this a problem? How can one understand this when they are significant? Is it unwise to include them?

All the best

[Felix Bittmann]

I think this is fine. As long as you derive the term from theory and it is significant, wonderful. Maybe the overall additional effect on R2 is very small but usually R2 is not of great relevance, especially if you are interested in causal mechanisms. On the other hand, if you goal is to maximize R2, the interaction term is probably not really relevant.

[George Hoffman]

i think this implies that the interaction changes the error in your model, but does not reduce it. you might want to look at how the other coefficients changed when interactions added?