probing interactions in quadratic multiple regression

[wale mobolaji]

I am trying to probe interactions between my predictor (X) and moderator (Z) in stata. where X has a quadratic relationship with the dependent variable.
I have used the margins command to test for slopes at different levels of Z (mean, 1 SD above mean and 1 SD below mean) but not sure if i am on the right track, as my results seem of

margins, dydx(X) at(Z=(-0.4978,-0.0680,0.3618))

[daniel klein]

We would need to see your regression command; this should be

regress depvar c.x##c.x##c.z ...

or the equivalent, that is, all lower order terms and the interaction should be in the model. Only then will margins give you the correct answers.

Best
Daniel