[R-sig-eco] GLM mixed model with quasibinomial family

[Kingsford Jones]

On Thu, Jul 29, 2010 at 10:07 AM, Javier Martinez
<javi.martinez.lopez at gmail.com> wrote:
> The question is, can I really use a
> mixed model if I only have two cases per group?

In general, yes.  E.g., the matched pairs t-test is subsumed by the LMM.

However, it sounds as though the plan is to fit many quasibinomial
GLMMs to a sample of
size somewhere between 8 and 16 (i.e. 8 if using site mean as the
response, 16 assuming independent observations, and somewhere in between for a
mixed model), and then assess significance using p-values; if this is
the case, my advice is to seek the help of a local statistician to
devise a different approach.


Kingsford