Skip to content

fixed observation-level variance in glmm

2 messages · Coe, Richard (ICRAF), Paul Buerkner

#
Dear mixed modellers

I have a data set classified by two factors A and B both of which are random. There is also an observation-level (residual)  random term that has a known variance,  defined by a known constant cv.  Since the dispersion parameter in a Gamma glm is the  cv, I want to do something like:

glmer(y~(1|A)+(1|B),  family=Gamma(link="identity"), dispersion = cv  ) where cv is known.

I can not see this facility or syntax in any of the mixed model tools I know of. Is there an easy way to fit this model?

Thanks
Ric

Richard Coe
Principal Scientist - Research Methods
World Agroforestry Centre (ICRAF), Nairobi, Kenya
and
Statistics for Sustainable Development, Reading, UK

Phone: +447734104196, +358503247733
#
If you are willing to go Bayesian, you can use the brms package. For your
model, the syntax would look as follows

brm(bf(y~(1|A)+(1|B), shape = cv), family = Gamma(link="identity"), ...)

Best,
Paul

2017-11-23 15:24 GMT+01:00 Coe, Richard (ICRAF) <R.COE at cgiar.org>: