[R-sig-ME] How to check if the interaction term is significant?

[Ben Bolker]

geetha venkatesh <geetha.r.v at ...> writes:

> 
> Dear lmer users,
> 
> I am using this package for the first time. I have a very basic question
> and I hope I can clarify my doubt. I would like to check if the interaction
> between two variables is significant. If I use two models, one with
> interaction term and the other without interaction and run a anova test,
> does this make sense?
> 
> m1=lmer(res~variant*age+sex+(1|batch))
> m2=lmer(res~variant+age+sex+(1|batch))
> 
> anova(m1,m2)
> 
> Is this the correct way to fit models in order to check for the
> significance of the interaction terms?
> 

  Yes; this runs a likelihood ratio test.  It is an asymptotic
test; you may want to consider the alternatives in the pbkrtest
package for other kinds of tests, but the basic idea is the same.

  Ben Bolker