[R-sig-ME] Estimating p-value for fixed effect which occurs also ininteraction

[David Duffy]

On Tue, 8 May 2012, Martin Heggli wrote:

> Dear listers,
>
> I am evaluating an experiment in plant ecology. Since the experiment has 
> both crossed and nested random factors and the data are unbalanced, I 
> decided to use linear mixed effects models (lmer from lme4). Right now, 
> I am struggling to determine the significance of the fixed factors. I 
> tried to estimate the p-value for the fixed factors by dropping 
> individual factors from the model and comparing the two models with 
> anova (cf. code below). However, I run into trouble when I try to drop 
> the factor Origin from the model which occurs also in the Origin:Soil 
> interaction term.

The main effect of Origin gets absorbed by Origin:Soil ie you have only 
changed the interpretation.  If Origin:Soil is necessary to the model, you 
probably want to examine Soil within each level of Origin.

Just 2c, David Duffy.