[R-sig-ME] heritability from longitudinal zero-inflated count data

[Jarrod Hadfield]

Not having a good morning!

random=~us(trait):animal+idh(at.level(trait,1):time):year

should have read

random=~us(trait):animal+idh(at.level(trait,1):year):nest

Jarrod




Quoting Jarrod Hadfield <j.hadfield at ed.ac.uk> on Fri, 03 Apr 2015  
08:27:01 +0100:

> Hi,
>
> It is possible, but you will need a lot of data. For example
>
> counts=~trait-1+x:trait,
> random=~us(trait):animal+idh(at.level(trait,1):time):year
>
> this models separate intercepts for the binary and count parts, and  
> different regressions on x. us(trait):animal estimates the additive  
> genetic variance in both parts, and the additive genetic variance  
> between them.  idh(at.level(trait,1):year):nest  fits different  
> between-nest variances for different years for the count part (trait  
> 1).
>
>
> Cheers,
>
> Jarrod
>
>
>
> Quoting Paul Johnson <paul.johnson at glasgow.ac.uk> on Thu, 2 Apr 2015  
> 17:03:47 +0000:
>
>> Hi all,
>>
>> I'd like to fit a model with the following features:
>>
>> * The data are parasite counts recorded at a number of time points  
>> (say 4) for each individual
>> * Zero-inflation, i.e. a mixture of binary and count data
>> * Separate fixed effects for the binary and count components
>> * Separate random effects for both components. One of the random  
>> effects will have a pedigree-derived correlation structure (as in  
>> an animal model). The random effects will have the same structure  
>> for both components, but need to allow different variances.
>> * The random effect variances are allowed to vary over time.
>>
>> The main aim is to estimate heritabilities for each time point,  
>> separately for the binary and count parts of the mixture, because  
>> the factors driving the two processes (encountering parasites and  
>> resistance to parasites) are expected to be driven by quite  
>> different factors, genetic and otherwise.
>>
>> Can this be done outside DIY software such as JAGS? I'd be  
>> interested in knowing how close MCMCglmm can get to this model,  
>> even if it can't do everything.
>>
>> Thanks for your help,
>> Paul
>>
>> _______________________________________________
>> 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.
>
> _______________________________________________
> 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.