[R-sig-ME] bivariate logistic/probit with hierarchical structure

[Jarrod Hadfield]

The `residual' variances would not be identifiable even though the  
residual correlation is, which I think would cause problems unless the  
variances are fixed at some value (may be possible with the  
development version of lmer? )


uoting David Duffy <davidD at qimr.edu.au> on Wed, 11 May 2011 16:10:04  
+1000 (EST):

> On Fri, 6 May 2011, Arthur Charpentier wrote:
>
>> I would like to fit a bivariate probit model, i.e. I observe Y=(Y1,Y2),
>> I have some covariates X1,...Xk, and I assume that there is a hierarchical
>> structure (and random effects), i.e. Y_i is actually a Y_{i,j} where j is a
>> subgroup index.
>> So this yield two correlations levels,
>> - within subgroup correlation between Y1_i's (if they belong to the same
>> subgroup)
>>
>
> Since you are doing "only" a bivariate, you might consider rewriting  
> it as a univariate problem with appropriate random effects (that is  
> Y1 and Y2 are in a lower level with variable specific random  
> intercepts etc).  This has been described previously sometime on  
> this list.
>
> Just 2c, David Duffy.
>
> -- 
> | David Duffy (MBBS PhD)                                         ,-_|\
> | email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
> | Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
> | 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v
>
> _______________________________________________
> 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.