[R-sig-ME] MCMCglmm: ixn between binary family-level factor and us() random effect
j.hadfield at ed.ac.uk
Wed Feb 9 11:16:36 CET 2011
Yes - the only issue with fixing the residual variances close to zero
is poor mixing, but if run for long enough its should be OK. However,
I would really check things are mixing by looking at the traces of all
effects. Also, I presumed that your response is Gaussian, but I
noticed in the subject header you use the word binary, and from
David's email it sounds like he believes the response is binary also.
If this is the case, the mother/father variances in the G-structure
cannot be estimated from the data either. It is usual to fix these at
1 so that the off-diagonals are correlations (which can be estimated).
This is going to be hard to set up in MCMCglmm and so Fisher might be
a better option (I'm not familiar with it).
On 9 Feb 2011, at 02:00, David Duffy wrote:
> On Wed, 9 Feb 2011, Paul Johnson wrote:
>> Hi Jarrod,
>> I've now compared the correlations estimated in MCMCglmm from a
>> similar model to the one below (4x4 family VCV in a single group,
>> mother and father residuals fixed to zero) with correlations
>> estimated using FISHER and they are similar enough to reassure me
>> that they're measuring the same thing.
> The advantage of FISHER is that it is "easy" [for a value of easy
> where easy=using FISHER ;)] to do an ascertainment corrected
> analysis (using asthma to indicate probands, and possibly asthma
> prevalence in the population to allow for your presumed Weinberg
> case-proband type design). Doing this is MCMC usually requires
> tricks, such as the "ones" trick.
> 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
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
More information about the R-sig-mixed-models