Can I drop the interaction term from the model if the interaction term is insignificant?
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
If the purpose of your analysis is to say something about the interaction term, then it is more convincing to keep the interaction in and show that it is not significant.
-
$\begingroup$ And by keeping it you will preserve type I error and unbiasedness of the estimate of residual variance. $\endgroup$ Commented Nov 25, 2014 at 13:09