[R-sig-ME] Family specification - MCMCglmm survival analysis

[Jarrod Hadfield]

Hi Steve,

"ordinal" has only been implemented for categorical data: you could  
expand each observation out into a series of zeros and ones and then  
fit an additional random effect pertaining to the original  
observation. However, that would be very inefficient computationally,  
and so if you're happy to use a logit link I would stick with  
"multinomial2".

Cheers,

Jarrod



probit link, "multinomial2" for logit link.


Quoting Steven Brady <steven.brady at yale.edu> on Sat, 3 Mar 2012  
13:54:28 -0500:

> Dear All:
> I have an experiment in which a suite of field enclosures were each  
> stocked with approximately 100 frog embryos. I would like to analyze  
> survival from this experiment as a two column response, e.g.  
> cbind(survived, died). Is it possible to use a probit link (i.e.  
> family = "ordinal") for a non-binary response such as this? If not,  
> would "multinomial2" be an appropriate family for this analysis? For  
> example:
> MCMCglmm(cbind(survived,(died))~x, random = ~clutch + block, family  
> = "multinomial2", data = xx)
>
> Many thanks,
> Steve
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>



-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.