Using the 'onlyuse' parameter in asreml.predict

I'm trying to understand just when I should use 'only' with predict(). My variance components look like this:
  
Code:
                  gamma component std.error z.ratio constraint
ped(RowTree) 5.965357e-01 9.543092e+01 1.293414e+02 0.73782176 Positive
Subblock 1.011929e-07 1.618835e-05 6.775276e-06 2.38932739 Boundary
Pplot 2.156411e-07 3.449723e-05 1.443805e-05 2.38932739 Boundary
Female 1.804322e-01 2.886467e+01 3.754475e+01 0.76880724 Positive
Male 1.296493e-02 2.074068e+00 2.238159e+01 0.09266845 Positive
FamilyNo 1.113177e-02 1.780807e+00 7.096272e+00 0.25094959 Positive
R!variance 1.000000e+00 1.599752e+02 6.695407e+01 2.38932739 Positive

We're interested in (among other things) what the relatively large number of trees identified by RowTree can tell us about the individual parents described in the pedigree. In case it's not obvious, each progeny is a member of a FamilyNo which is related to the Male and Female columns in a way described in the pedigree.

If I do a prediction with no 'only' term, I end up with heritability for the parents I'm interested in of something like 0.68 (steps not shown). If I include a term
Code:
only = 'ped(RowTree)'

, it's about 0.59.

What I need to know is under what circumstances would it be appropriate to use and when should I omit the 'only' argument? If I understand correctly, the predict stage will be averaging over the other terms. I can't see just when that would make sense. Perhaps it's just that it's not doing what I think it does, so I'd be grateful to receive any clarification.

Would it be different if I'm calculating heritability of the progeny?

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

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

Read this topic online here:
http://www.vsni.co.uk/forum/viewtopic.php?p=511#511

-------------------- m2f --------------------
Received on Wed Apr 28 2009 - 00:55:03 EST