Re: heterogeneous error structure

Hi ASreml users,

I sent this message to Dr Arthur Gilmour a few days ago. He is out can not
respond before Aug. 21. I have two questions
1. How to fit error variance as a quadratic or other polynomial function, etc?
2. After fitting c spline to DIM how to get the var-covariance on daily basis. I guess I
have to use the Z matrix in *.res file.
When I fit polynomials I use the coefficients (Q) from the *.res file and
pre and post multiple the variance-covariance (G).
Ie Q'GQ.

Mekonnen
----- Forwarded by Mekonnen HaileMariam/NRE on 08/08/2006 04:13 PM -----

Mekonnen HaileMariam
07/08/2006 11:23 AM

 
        To: arthur.gilmour_at_AGRIC.NSW.GOV.AU
        cc:
        Subject: Re: heterogeneous error structure

Dear Arthur Gilmour,

Dear Arthur Gilmour,
This is a response you sent to the question that was raised about fitting
heterogeneous error structure.
I am trying to analyse test day milk yield data recorded over 18 to 24
months of period so there are about 18 to 24 independent error variances.
Instead of fitting 18 or 20 error variance I want to fit the residual
variance (e.g. as a quadratic or other polynomial function), such that Vr
is a continuous function of days in milk.

When I try to run the as file examples that you sent to the list I am
getting error with G or R header lines.
I could run the situation with homogenous error variance.

My *.as file as below

Random regession analysis of milk yield
 cow 10000 !I
 sire !P
  ys 13 !I
 herdTD 31083 !I
 ageAtTest
 milkY !M0 !*0.1
 DIM
 TestNum
sirePed
TestDATA !MAXIT 20 !ASUV
milkY ~ mu pol(DIM,-6) pol(ageAtTest,-3) !r pol(DIM,5).cow !r
pol(DIM,4).sire ,
 !f herdTD mv
1 2 2 !STEP .01
10000 !S2==1
18 0 DIAG 5.96 !GP +17
 5.90
 5.96
 5.96
 5.96
 5.90
 5.96
 5.96
 5.96
 5.90
 5.96
 5.96
 5.96
 5.90
 4.96
 4.96
 4.96
 4.90
pol(DIM,1).cow 2
pol(DIM,1) 0 US 8.974 !+2 !GP
   4.012 14.87
   2.033 5.662 9.557
  -0.2175 0.1828 3.151 7.860
   0.02175 0.01828 0.3151 0.7860 2.1
   0.002175 0.001828 0.03151 0.07860 0.21 0.11
 cow

pol(DIM,4).sire 2
pol(DIM,4) 0 US 2.974 !+14 !GP
   1.012 4.87
   1.033 1.662 2.557
   0.02175 0.1828 1.151 2.860
   0.002175 0.01828 0.01151 0.2860 0.25
   sire

After this I want to try splines instead of pol.

Thanks for your help
Mekonnen Haile-Mariam

 
----- Forwarded by Mekonnen HaileMariam/NRE on 07/08/2006 09:57 AM -----

arthur.gilmour_at_AGRIC.NSW.GOV.AU
Sent by: ASREML-L_at_AGRIC.NSW.GOV.AU
09/03/2005 12:26 PM
Please respond to ASREML-L

 
        To: ASREML-L_at_AGRIC.NSW.GOV.AU
        cc:
        Subject: Re: heterogeneous error structure

Dear Alistair,

I enclose a random regression example for your consideration.

Following is an amended version of your job.

TRAIT ~ mu SEX AGEF !r pol(sAGE,2).ANIMAL pol(sAGE,2).ide(ANIMAL) MOTHER
 !f mv
0 0 2
#1 1 3

#0 0 ID 3 # single homogeneous error variance

pol(sAGE,2).ANIMAL 2
pol(sAGE,2) 0 US !GP
1.64
0.96 7.801
0.01 0.01 0.01
ANIMAL 0 AINV

pol(sAGE,2).ide(ANIMAL) 2
pol(sAGE,2) 0 US !GP
1.64
0.96 7.801
0.01 0.01 0.01
ide(ANIMAL) 0 ID

#MOTHER 1
#MOTHER 0 IDV 0.226

This assumes an animal model with repeat observations in time
for the animals. This has a common independent residual variance plus
the animal level random regression covariance structure at the residual
and the genetic level.

You could extend it to have a random regression for the maternal
component.

It is sometimes difficult to get solutions for the US variance structures
as it is difficult to identify the good starting values. !GP should
help.

Best wishes in the Name of Jesus Christ, my Saviour.
Check out http://members.dodo.net.au/~cargovale

Arthur Gilmour PhD
Principal Research Scientist (Biometrics)
NSW Department of Primary Industries
Orange Agricultural Institute, Forest Rd, ORANGE, 2800, AUSTRALIA

mailto:Arthur.Gilmour_at_agric.nsw.gov.au
fax: 02 6391 3899; 02 6391 3922 Australia +61
telephone work: 02 6391 3815; home: 02 6364 3288;

http://www.agric.nsw.gov.au/reader/17263
http://www.asreml.co.uk/
The ASReml discussion group ASREML-L_at_agric.nsw.gov.au
   has reactivated. To join, mailto:arthur.gilmour_at_agric.nsw.gov.au
Cookbook: http://uncronopio.org/luis/asreml_cookbook.html

Ellen is doing well after her Hip replacement on 24 Feb but
she will need crutches for a while yet.
<><><><><><><><><><><><><><><><><><><><><><><><><>

This message is intended for the addressee named and may contain
confidential information. If you are not the intended recipient or
received it in error, please delete the message and notify sender. Views
expressed are those of the individual sender and are not necessarily the
views of their organisation.