Interpreting interaction term when the main effect sign flips

[Linus Seb]

Hi, this is a fairly basic question, I added a control variable to my fixed effect model and the sign of one of the main variables flipped and became negative and the sign of the interaction term flipped and has become positive. In general, when the sign of the main effect is negative and the interaction term is positive, can I still interpret the interaction term as the conditional effect of a unit increase in the other term (not the main effect). In other words, if my model is

Y=B₀ + B₁X₁ - B₂X₂ +B₃X₁*X₂

Can I interpret B₃ as X₂ increaes Y by B₃ conditional on a one unit increase in X₁?

Is there anything in particular to be aware of when the main effect is negative but the interaction is positive?

[Richard Williams]

https://www3.nd.edu/~rwilliam/stats2/l53.pdf

That handout talks about interpreting main and interaction effects. The key thing to realize is that the interpretation of main effects changes greatly once you add interactions. Things like sign flips May not mean much. It also lists things that can make interpreting easier, like centering continuous variables before interacting.

[Carlo Lazzaro]

Linus:
as an aside to Richard's helpful advice and link to his really valuable handout, please follow the FAQ and post what you typed and what Stata gave you back. Thanks.