I am attempting to analyze the effect of two categorical variables (
species) on a continuous variable (
carbon) though a linear mixed model analysis. Study sites are included as the random effect in the model (with the random slope and random intercept). Landuse, species (and their interaction) are included as fixed effects.
the model is this -
model1 = lmer(carbon ~ species*landuse + (1+landuse|site), data)
I know that there may be interaction between landuse and species. I know that presence of interaction can change the interpretation of the main effects. I want to know what should I do if there is a significant interaction between landuse and species? In that case, do I study the effect of landuse for each species seprately with the following model -
model.sp1 = lmer(carbon ~ landuse + (1+landuse|site), data.sp1)
and repeat this for all the six species? Do I need any form of corrections (of p-values) due to running multiple tests?
Another question is that if the interaction term is not significant, can I interpret the main effects from
model1 or do I run another model (
model2) without the interaction effect and interpret the main effects from there?
model2 = lmer(carbon ~ species + landuse + (1+landuse|site), data)
I am fairly new to mixed model and R, so please excuse my naivety!
PS - just to clarify, I have a fairly good idea of what an interaction mean and how to interpret it. I do not know, how to interpret main effects in the presence of an interaction - whether I need to run a seprate analysis to interpret main effects.