[R-sig-ME] variance estimates of main effects change dramatically depending on interactions permitted

[Reinhold Kliegl]

Is there a significant improvement in goodness of fit due to
estimating the variance for the interaction effect (i.e., anova(m1,
m2)?
Are V1 and V2 factors (what kind of contrasts?) or covariates? Are V1
and V2 correlated? N of observations at various levels?
Reinhold Kliegl

On Fri, Oct 21, 2011 at 7:55 PM, Mike Lawrence <Mike.Lawrence at dal.ca> wrote:
> I am interested in estimating the variance in the effect V1 across
> levels of my random effect ("id"). I also happen to have another
> variable, V2, crossed with V1. The model:
>
> lmer(
>        data = my_data
>        , formula = dv ~ V1*V2 + ( V1 + V2 | id )
> )
>
> yields a random effects variance estimate for V1 of about 10. However,
> the model:
>
> lmer(
>        data = my_data
>        , formula = dv ~ V1*V2 + ( V1*V2 | id )
> )
>
> yields a random effects variance estimate for V1 of about 100.
>
> Any idea what would cause such a large difference, and which variance
> estimate is more appropriate?
>
> Cheers,
>
> Mike
>
> --
> Mike Lawrence
> Graduate Student
> Department of Psychology
> Dalhousie University
>
> Looking to arrange a meeting? Check my public calendar:
> http://goo.gl/BYH99
>
> ~ Certainty is folly... I think. ~
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>