[R-sig-ME] nested random effect specification in MCMCglmm

[Jarrod Hadfield]

Hi Rebecca,

If every nest has a unique identifier then MCMCglmm should give the  
same answer (up to Monte Carlo error) for random ~ pairID + nest and  
random ~ pairID + pairID:nest, so a difference is worrying. Can the  
difference for the significance be explained by Monte Carlo error?  
Perhaps you could post the summaries from the two models?

Cheers,

Jarrod


Quoting "Sardell, Rebecca" <rebecca.sardell at abdn.ac.uk>:

> Hi,
>
> I'm using MCMCglmm to run some binary and Poisson models and I'd   
> just like to check whether I'm specifying a nested random effect   
> correctly.
>
> I have 773 data points, each one corresponding to a chick in a nest.  
>  I want to test for differences between half-siblings so I'm using   
> three random effects: natal year, nest and pairID.  There are 17   
> natal years so year was included as a blocking factor.  There are   
> 245 nests each with a unique number, and 177 different parent pairs,  
>  each with a unique number.  Some parent pairs have >1 nest so nest   
> is nested within pairID.
>
> My understanding is that in lmer (1|pairID/nest) is equivalent to   
> (1|pairID) + (1|pairID:nest) which is equivalent to (1|pairID) +   
> (1|nest) as long as each level of nest has a unique value, which it   
> does. Using my dataset in a binary model I get the same results for   
> each of the above in lmer so that's fine.
>
> I'm just wondering whether this is the same when specifying nested   
> random effects in MCMCglmm? I'm guessing it's not as specifying   
> random ~ pairID + nest compared to random ~ pairID + pairID:nest in   
> the model below gives me different significance levels for one of my  
>  main effects.  Comparing these results with the binary model in  
> lmer  suggests that I should probably be using ~ pairID +  
> pairID:nest when  using MCMCglmm but I'm not completely sure. Is  
> this correct or  should I be able to use either??
>
> priorX1 = list(R = list(V = 1, n = 0, fix = 1), G = list(G1 = list(V  
>  = 1, n = 0.002), G2 = list(V = 1, n = 0.002)))
> modelX1 <- MCMCglmm(y ~ C + D + C:D, random = ~ natalyr + pairID +   
> pairID:nest, family = "categorical", data =early, prior = priorX1,   
> burnin = 3000, nitt = 1003000, thin=1000)
>
>
> Thanks,
>
>
> Rebecca Sardell
> PhD Student
> Institute of Biological & Environmental Sciences
> University of Aberdeen
> Zoology Building
> Tillydrone Avenue
> Aberdeen
> AB24 2TZ
> Scotland
>
>
>
> The University of Aberdeen is a charity registered in Scotland, No SC013683.
>
> _______________________________________________
> 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.