Possible Duplicate:
Interpreting main effect and interaction

In a GLM, I had one categorical predictor and one covariate. To ensure the covariate was consistent across levels of my predictor I included an interaction term. This was non significant. Am I then justified in removing it from the model? My categorical predictor is non significant when it is included, but significant when it is removed.

Any help would be much appreciated.


marked as duplicate by whuber Jul 26 '12 at 11:52

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • $\begingroup$ Does any underlying theory predict the interaction ought to be there? $\endgroup$ – whuber Jul 26 '12 at 0:13
  • $\begingroup$ Not that I am aware of. I included it only to test the assumption of homogeneity of regression slopes. $\endgroup$ – Heather Jul 26 '12 at 7:40
  • $\begingroup$ Then I'm sure your question is a duplicate of several closely related questions about regression, significance, and covariates, including (but not limited to) stats.stackexchange.com/questions/19891 and stats.stackexchange.com/questions/22680. For more, search our site. $\endgroup$ – whuber Jul 26 '12 at 11:51