Re: Binary trait random regression

From: Bruce Southey <southey_at_UIUC.EDU>
Date: Tue, 22 Jan 2008 10:43:23 -0600

If you can find some archives you will get some answers :-)

The simple answer to using generalized linear mixed animal models is that it depends on the data and the method. For example, Moreno et al.(1997 Genet. Sel. Evol.) showed that the maximum a posteriori approach provides biased inferences. On this list Arthur has indicated bias by the approach in ASREML but I do not know any other details. However, I have not seen that in analysis of real data with either ASREML or Matvec (rather bias by not including all sources of variation - like maternal effects).

I think that most of these issues are probably not relevant because you have repeated observations per subject.

You can use the deviance to compare models. You should also look at an estimate of dispersion as it can be a good indicator of model fit.


---- Original message ----
>Date: Tue, 22 Jan 2008 15:36:08 +0000
>From: Alastair Wilson <Alastair.Wilson_at_ED.AC.UK>
>Subject: Binary trait random regression
>Hello all,
>I have a question regarding the wisdom of fitting a random regression
>model of fecundity (scored as 0 or 1) in a human data set where
>fecundity has been scored annually for several hundred women for each
>year of their life (from 16-50 I think). I have fitted the two models
>below and get apparently sensible results from the second of these in
>which I wanted to test for a genotype by age interaction.
>fec !binary ~ mu fixedFX !r leg(ageS,0).ANIMAL leg(ageS,0).ide(ANIMAL)
>fec !binary ~ mu fixedFX !r leg(ageS,1).ANIMAL leg(ageS,1).ide(ANIMAL
>However, I'm also aware that there are issues relating to the use of
>generalised animal models, and was wondering if anyone had views on
>whether this was valid? Also, if the approach is not inherently wrong,
>is there any way to statistically compare these models if, as I
>understand, the likelihoods should not be used for model comparison.
>Best wishes,
>Alastair J Wilson
>NERC Research Fellow
>tel. 0131 6513608
>Institute of Evolutionary Biology
>School of Biological Sciences
>The University of Edinburgh
>Ashworth Laboratories
>The King's Buildings
>West Mains Road
>Edinburgh EH9 3JT, UK
Received on Fri Jan 22 2008 - 10:43:23 EST

This webpage is part of the ASReml-l discussion list archives 2004-2010. More information on ASReml can be found at the VSN website. This discussion list is now deprecated - please use the VSN forum for discussion on ASReml. (These online archives were generated using the hypermail package.)