[R-sig-ME] glmmADMB without random effects

[Alexej Siren]

Hello,

 

I have a question about differences in model output.  If random effects are
not specified for a model using the glmmadmb function then the parameter
estimates should be the same for glm, correct?  Out of curiosity I tested
this on several datasets and found that the magnitude, but not the
direction, of parameter estimates changed for some models.  The only pattern
that I detected that was that this happened for datasets that contain highly
overdispersed data.  Is there a difference in fitting algorithms that would
cause this pattern?  Any insight would be appreciated.  

 

If there is not a straightforward explanation I can provide data.  

 

Thanks,

 

Alex 

 

 


	[[alternative HTML version deleted]]