Re: true BLUP (fwd)
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: true BLUP (fwd)



Thanks to Dr. Gilmour, Apiolaza adn Dutkowski for your help.
The following is the answer by Dr. Gilmour. Sorry I did not forward
before. I was out of town these days.

I tried Dr. Gilmour suggestions. They work as he said.
Thanks again.
Uilson


---------- Forwarded message ----------
Date: Mon, 20 Jul 1998 08:01:58 +1000 (EST)
From: Arthur Gilmour <gilmoua@ornsun.agric.nsw.gov.au>
To: wilsonvl@grove.ufl.edu
Subject: Re: true BLUP

Dear Uilson,

There are two points to understand.

1  When using the !GF option, it must be specified for every parameter.
   Thus the job would be like
   
  True BLUP
  env            2  #  environments
  fam           80  # families
  blk           12  # blocks
  x1
 /f/share/uvl/CHAP3/BLUPtrue/CH3e.dat  !maxit 1
  x1  ~ mu blk env !r fam 0.0013468 !GF fam.env 0.052304 !GF blk.fam 0.11188 !GF
  1 1 0
  0 !S2==0.012249  # !GF
  
  
If you look at the results below, you will see that the Error variance was 
fixed in BOTH (!S2==  fixes it).
The !GF is not needed on that line because there are no other parameters for it
to apply to.

Another way of writing the last line is
 0 0 0 0.012249 !GF !S2==1
 

The blk.fam variance was also fixed in that second run 
    because the !GF followed 0.11188

> 

2.  Both analyses are giving you the 'true blup'
    BLUP referes to the character of the random effects.
    The random effects reported after the first iteration
    are calculated using the parameter values used in that iteration,
    the once that were input.  ASREML then provides REML updates of
    those parameters and would use the updated parameters in the
    second iteration.  The !GF procedure therefore alters
    the REML part of the process but not the BLUP part.
    I.e. The numbers in the .sln file are not altered by the
    presence or absence of !GF in the .as file [after the first
    iteration].  This comment may not apply to the Variance parameter
    fixed by the !S2== clause in that if you used !S2= instead, the
    SEs of the BLUPS might change in the .sln file.  You could confirm
    that.
    
 I trust this helps
 
 PS  You would need to add the !GF qualifiers to Luis version of your job to
 fix the parameters.  If you succeed in fixing the parameters, you could omit 
the !MAXIT 1
 and allow ASREML to do 2 iterations when it would stop because there is no 
change in the LogL.
 
 
 Arthur   
> From: Uilson V Lopes <wilsonvl@grove.ufl.edu>
> Comments: PAA01154 on bay (hop 0), Fri, 17 Jul 1998 15:44:30 -0400 (EDT)
> Date: Fri, 17 Jul 1998 15:44:30 -0400 (EDT)
> X-Sender: wilsonvl@bay
> To: asreml@ram.chiswick.anprod.csiro.au
> Subject: true BLUP
> MIME-Version: 1.0
> 
> Dear asremlrs,
> Sorry to return to the TRUE BLUP problem.
> 
> I am trying to run true BLUP in a simulated data set. 
> 
> I tried the option !maxit, but the final variance components are different
> from my priors (Analysis 1, below).
> 
> Then I tried to add the option !GF, which was supposed to keep fixed the
> variance components. Again, the estimates of the variance components are
> different from my priors (Analysis 2, below).
> 
> Could somebody help me solve this?
> Thanks
> Uilson Lopes
> 
> 
> 
> ====================================
> .as file
> 
>  True BLUP
>  env            2  #  environments
>  fam           80  # families
>  blk           12  # blocks
>  x1
> /f/share/uvl/CHAP3/BLUPtrue/CH3e.dat  !maxit 1
>  x1  ~ mu blk env !r fam 0.0013468 fam.env 0.052304 blk.fam 0.11188  # !GF
>  1 1 0
>  0 !S2==0.012249  # !GF
> 
> ======================================================
>  ANALYSIS 1. NO !GF option
> 
> 
>  LogL=-285885.     S2= 0.1225E-01  9588 df   0.00135   0.05230   0.11188
>      0.01225
>  Final parameter values                      0.00549   0.06940   0.14017
>      0.01225
> 
>    Source       Model  terms     Gamma     Component    Compnt/StndErr
>   fam              80     80  0.549387E-02  0.549387E-02      0.84
>   fam.env         160    160  0.694019E-01  0.694019E-01      7.33
>   blk.fam         960    960  0.140167      0.140167         28.44
>   Variance       9600   9588  0.122490E-01  0.122490E-01      0.00
> =====================================================
> ANALYSIS 2. WITH !GF option
>  
>  LogL=-285885.     S2= 0.1225E-01  9588 df   0.00135   0.05230   0.11188
>      0.01225
>  Final parameter values                      0.00549   0.07412   0.11188
>      0.01225
> 
>    Source       Model  terms     Gamma     Component    Compnt/StndErr
>   fam              80     80  0.549387E-02  0.549387E-02      0.84
>   fam.env         160    160  0.741164E-01  0.741164E-01      7.86
>   blk.fam         960    960  0.111880      0.111880          0.00
>   Variance       9600   9588  0.122490E-01  0.122490E-01      0.00
> 
> ============================================================
> 


<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD                    email: Arthur.Gilmour@agric.nsw.gov.au
Senior Research Scientist (Biometrics)                 fax: <61> 2 6391 3899
NSW Agriculture                                             <61> 2 6391 3922
Orange Agricultural Institute               telephone work: <61> 2 6391 3815
Forest Rd, ORANGE, 2800, AUSTRALIA                    home: <61> 2 6362 0046

ASREML is currently free by anonymous ftp from pub/aar on ftp.res.bbsrc.ac.uk
    Point your web browser at ftp://ftp.res.bbsrc.ac.uk/pub/aar/ 
    in the IACR-Rothamsted information system  http://www.res.bbsrc.ac.uk/ 

To join the asreml discussion list, send the message  
     subscribe
to  asreml-request@chiswick.anprod.CSIRO.au

The address for messages to the list is asreml@chiswick.anprod.CSIRO.au

                        <> <> <> <> <> <> <>
"The men marveled, saying 'Who can this be,
         that even the winds and the sea obey Him?'"  Matthew 8:27
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>