Re: Random Regression - maternal effects

Dear Boyd,

It is always difficult to diagnose convergence problems without seeing the whole job.

Now, first I would be adding error covariance to the model

leg(age,1).ide(ID)

as it is not unlikely that the variance changes with age (depending on the trait)
and your model does not allow that at the error level.

IF AI singularity occurs in the first iteration, that indicates there is a
structurral problem with the data/model. That is, the structure of the model does not allow the model to be fitted because of lack of information to estimate all the parameters.

I gather in your case, it is failing to converge and sometimes ending up with silly parameter values resulting in a singularity in the AI matrix.

This could simply be that the model is not appropriate and the aforementioned
omission of leg(age,1).ide(ID) could be enough to cause the problem.

So I suggest you try
!PART 4 # random regression on both direct and dam effects
wgt ~ mu leg(age,1) bt eweage sex !r leg(age,1).ide(ID),
 !{ leg(age,1).ID leg(age,1).Dam !} !f cg
1 1 3
26160 # number of individuals in dataset

leg(age,1).ide(ID) 2
2 0 US !GP
 50 30 30
ide(ID)

leg(age,1).ID 2
2 0 US !GP
 50 30 30
ID 0 AINV

leg(age,1).Dam 2
2 0 US !GP
 10 5 5
Dam 0 AINV

And if you get that running, and there is slope variation
associated with Dam,

Then try adding back the genetic correlations between ID and Dam.

If there is no maternal effect on the age regression (and it is not likely there is),
then that could be the reason the more general model is failing.

------------------------
Arthur Gilmour

Principal Research Scientist (Biometrics)
NSW Dept Primary Industries

-------------------- m2f --------------------

Sent using Mail2Forum (http://www.mail2forum.com).

Read this topic online here:
http://www.vsni.co.uk/forum/viewtopic.php?p=430#430

-------------------- m2f --------------------
Received on Wed Mar 24 2009 - 04:17:07 EST