Advice for analysis of biological data - Mixed model or NESTED-Anova?

Savani,
I think your problem is that your model is using group as both fixed and
random effect.
Try this instead:
So for your experiments with NO treatment, you need a model to assess the
effect of group, if you want to say, "Did the groups differ":
simple_model<-nlme(logVolume ~ Group , random= ~Animal_ID, data=savanidata)

For your experiments WITH a treatment,
treatment_model<-nlme(logVolume ~ Treatment , random= ~Group/Animal_ID,
data=savanidata)

If you actually want to know the interaction between treatment and group,
then you must have "Group" as fixed effect:
interaction_model<-nlme(logVolume ~ Treatment*Group , random= ~Animal_ID,
data=savanidata)
Then you can test the terms like this:

model1<- update(interaction_model, ~. - Treatment:Group) #excludes
interaction

anova(interaction_model, model1) #likelihood ratio test; if p<0.05, the
interaction term is doing a good job in your model, so you should keep the
interaction term
Otherwise, you can keep simplifying, like this:
model2<-update(model1, ~. - Group)
anova(model1, model2)
And so you can test the importance of each term.

Happy modeling,
Evan

On Thu, May 26, 2016 at 5:17 AM, Savani Anbalagan <savani1987 at gmail.com>
wrote:

Advice for analysis of biological data - Mixed model or NESTED-Anova?

Thread (4 messages)