I have a model that has two significant main effects and where the interaction between the two variables (age and %gray) is also significant in a multiple regression. When the interaction (age * %gray) is present in the model, the beta of age changes sign with respect to the model where the main effects alone are included. Is it possible to plot the data (independent variable vs. dependent) in a way that reflects the new sign of the main effect, after the interaction term has been included?

  • 2
    $\begingroup$ Yes, you can consider plotting a 3-D surface (x=age, z=gray, y=outcome), or you can consider binning gray by, say, quintiles and then proceed to make five smaller outcome vs. age scatter plots to show the changes in slope as you move up in the %gray quintile. Of course, you can overlay the five regression lines onto the same plot as long as they don't look messy. $\endgroup$ – Penguin_Knight Mar 20 '14 at 21:08

Your Answer

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

Browse other questions tagged or ask your own question.