The extra 'Variance' term seems to be the result of using !BLUP
I tried various bits of alternative code (without the maternal and litter
effects to make things simpler). I always got the 'Variance' term using
!BLUP, but did not get it when I used !maxit 1 instead.
One thing I noticed using !maxit 1 was that the variance components get
updated, but the BVs must be estimated before that happens, as I get
identical BVs whether using BLUP or maxit 1, and also when using maxit 1
with !GF on the variance components (in which case they do not get updated).
The SEs of the BVs were the same for the two maxit 1 runs, but were
different from using BLUP 3. Maybe the 'Variance' term is getting in the
way.
Has anyone else had any experience with these matters?
Thanks,
Ken Dodds.
-----Original Message-----
From: Dodds, Ken [mailto:ken.dodds@agresearch.co.nz]
Sent: Friday, April 06, 2001 5:55 PM
To: ASREML message (E-mail)
Subject: Bivariate analysis with maternal effects
I am trying to do a bivariate analysis with maternal effects for one of the
traits. I have done this previously (a few years ago) with another data set,
but now I seem to be getting an extra 'Variance' variance component, with or
without maternal effects correlations specified, and trying a variety of
ways of specifying the components. I would be grateful of any help. Thanks.
_____________________________________________________________________
Ken Dodds
AgResearch
Invermay Agricultural Centre
Private Bag 50034
Mosgiel
New Zealand
Ph: +64 3489-9083
Fax: +64 3489-9037
Email: ken.dodds@agresearch.co.nz
_____________________________________________________________________
Example models and variance components output.
The following are with !BLUP specified (i.e. should give initial parameters
back).
wwt lw8 ~ Trait Tr.sex Tr.br.rr Tr.aod Tr.bday Tr.yrflk !r Tr.id
at(Tra,1).did 0.5 at(Tra,1).litter 0.7
1 2 1
0 0 0
Trait 0 US 5.1 6.2 15 # residuals
Tr.id 2
Trait 0 US 4.6 5.7 10 # direct
id 0 0
gives
Source Model terms Gamma Component Comp/SE % C
at(Tra,1).did 11098 11098 0.500000 0.500000 0.05 0 P
at(Tra,1).litter 61986 3477 0.700000 0.700000 0.06 0 P
Variance 14472 14386 1.00000 1.00000 0.01 0 F
Residual UnStruct 1 5.10000 5.10000 0.42 0 U
Residual UnStruct 1 6.20000 6.20000 0.52 0 U
Residual UnStruct 2 15.0000 15.0000 3.78 0 U
Tr.id UnStruct 1 4.60000 4.60000 0.52 0 U
Tr.id UnStruct 1 5.70000 5.70000 0.66 0 U
Tr.id UnStruct 2 10.0000 10.0000 3.36 0 U
wwt lw8 ~ Trait Tr.sex Tr.br.rr Tr.aod Tr.bday Tr.yrflk !r Tr.id
at(Tra,1).did at(Tra,1).litter 0.7
1 2 1
0 0 0
Trait 0 US 5.1 6.2 15 # residuals
Tr.id 2
3 0 US 4.6 5.7 10 0 0 2.5 # direct + Maternal
id 0 0
gives
Source Model terms Gamma Component Comp/SE % C
at(Tra,1).litter 61986 3477 0.700000 0.700000 0.11 0 P
Variance 14472 14386 1.00000 1.00000 0.01 0 F
Residual UnStruct 1 5.10000 5.10000 0.48 0 U
Residual UnStruct 1 6.20000 6.20000 0.60 0 U
Residual UnStruct 2 15.0000 15.0000 4.21 0 U
Tr.id UnStruct 1 4.60000 4.60000 0.68 0 U
Tr.id UnStruct 1 5.70000 5.70000 0.86 0 U
Tr.id UnStruct 2 10.0000 10.0000 3.99 0 U
Tr.id UnStruct 1 0.00000 0.00000 0.00 0 U
Tr.id UnStruct 2 0.00000 0.00000 0.00 0 U
Tr.id UnStruct 3 2.50000 2.50000 0.41 0 U
wwt lw8 ~ Trait Tr.sex Tr.br.rr Tr.aod Tr.bday Tr.yrflk !r Tr.id
at(Tra,1).did at(Tra,1).litter 0.7
1 2 2
0 0 0
Trait 0 US 5.1 6.2 15 # residuals
Tr.id 2
Trait 0 US 4.6 5.7 10 # direct
id 0 0
at(Tra,1).did 2
1 0 US 2.5 # Maternal
id 0 0
gives
Source Model terms Gamma Component Comp/SE % C
at(Tra,1).litter 61986 3477 0.700000 0.700000 0.11 0 P
Variance 14472 14386 1.00000 1.00000 0.01 0 F
Residual UnStruct 1 5.10000 5.10000 0.48 0 U
Residual UnStruct 1 6.20000 6.20000 0.60 0 U
Residual UnStruct 2 15.0000 15.0000 4.21 0 U
Tr.id UnStruct 1 4.60000 4.60000 0.68 0 U
Tr.id UnStruct 1 5.70000 5.70000 0.86 0 U
Tr.id UnStruct 2 10.0000 10.0000 3.99 0 U
at(Tra,1).did identity 1 2.50000 2.50000 0.41 0 U
wwt lw8 ~ Trait Tr.sex Tr.br.rr Tr.aod Tr.bday Tr.yrflk !r Tr.id
at(Tra,1).did at(Tra,1).litter 0.7
1 2 3
0 0 0
Trait 0 US 5.1 6.2 15 # residuals
Tr.id 2
Trait 0 US 4.6 5.7 10 # direct
id 0 0
at(Tra,1).did 2
1 0 US 2.5 # Maternal
id 0 0
at(Tra,1).litter 2
1 0 US 3.6
litter 0 0
gives
Source Model terms Gamma Component Comp/SE % C
Variance 14472 14386 1.00000 1.00000 0.01 0 F
Residual UnStruct 1 5.10000 5.10000 0.51 0 U
Residual UnStruct 1 6.20000 6.20000 0.64 0 U
Residual UnStruct 2 15.0000 15.0000 4.42 0 U
Tr.id UnStruct 1 4.60000 4.60000 0.79 0 U
Tr.id UnStruct 1 5.70000 5.70000 0.99 0 U
Tr.id UnStruct 2 10.0000 10.0000 4.38 0 U
at(Tra,1).did identity 1 2.50000 2.50000 0.70 0 U
at(Tra,1).litter identity 1 3.60000 3.60000 1.01 0 U
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml