quasi-poisson

[Quresh Latif]

Has anyone tried implementing quasi-poisson regression in NIMBLE as described in the attached? Would the correct approach for implementing this be via NIMBLE function(s). If so, has anyone written up the quasi-poisson for NIMBLE, or can anyone point me to an example set of functions for defining a distribution that's not already supported by NIMBLE?

[dirkdouw...@gmail.com]

You could reparametrize the negative binomial. 

If y~ dnegbin(prob = r/(1+r), size = r*u) then E[y]=u and V[y]=u*(1+1/r)

[Quresh Latif]

Brilliant! I actually ended up solving the problem I was trying to deal with via another route, but hopefully it’s helpful to someone (possibly my future self) to have this solution in the ether.

 

Thank you!

 

Quresh S. Latif 
Research Scientist
Bird Conservancy of the Rockies
Phone: (970) 482-1707 ext. 15
www.birdconservancy.org

--
You received this message because you are subscribed to a topic in the Google Groups "nimble-users" group.
To unsubscribe from this topic, visit https://link.edgepilot.com/s/33134a1b/m09jBYnfekKt71nSuh9cfw?u=https://groups.google.com/d/topic/nimble-users/RSE6suHxFe8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to nimble-users...@googlegroups.com.
To view this discussion on the web visit https://link.edgepilot.com/s/5067a6bf/gldyUI2a_EawflFAkpOPfA?u=https://groups.google.com/d/msgid/nimble-users/3ae5750c-0d3b-4e9c-93a9-dd446e5eebc2n%2540googlegroups.com.