Skip to content
Prev 11473 / 20628 Next

Huber-White standard errors for glmer Poisson fits

I just tried your new weights.merMod(). 
Specifying "working" weights and "working" residuals gets me the exact same 
standard errors as with normal weights (all equal to 1) and "score" residuals, 
i.e. very close to what GLLAMM produces.
So I guess that was indeed the missing link - having a "working" option for 
weights.merMod(). Now one can easily extend estfun.glm() to work with merMod 
objects.

Thanks to both of you!

Theodore

???? ??? 17 ??? 2014 18:58:47 Steve Walker ??????: