Dear All,
I have just updated ASREML on BBSRC. The new version is slightly faster
on LARGE MET analyses and allows an AR.AR G structure to work.
I am working on a few other gremlins so unless doing MET's you may
like to wait before upgrading. For those doing METs, the following
works:
at(site,3).(row col col.lr)
as shorthand for
at(site,3).row at(site,3).col at(site,3).col.lr
Dear Joao
ASREML detects a singularity in the matrix
535 0.000001
0.000001 0.000001
But apparantly not in
0.000001 535
0.000001 0.000001
I believe you are using this structure because the trait
with variance 0.000001 is entirely missing at this site.
So, in the absense of information, you have put the value
close to 0.0. You could also reasonably use a value
of 4 which is closeer to the value for the other site where there is data.
ASREML needs to detect singularities but sometimes gets it wrong
as in this case. The reason it gets it wrong is related to
the big difference in scale between the traits. It uses
the first element to define a value for testing for a singulariity.
In the first case, the test value is 535 x 0.0000001 (maybe) = .0000535
which is bigger than 0.000001 so a singularity is reported.
In the second case, the test value is 1E-12 in which case
no singularirty is reported.
For this reason, it is helpful to the algorithm if
the traits have similar order of magnitude in the variance
or the one with smallest variance is nominated first.
Arthur
> Date: Tue, 16 Feb 1999 19:23:51 +0100
> From: Joao Costa e Silva <jces@kvl.dk>
> To: gilmoua@ornsun.agric.nsw.gov.au
> Subject: RE: Error message and some questions
> Mime-Version: 1.0
>
> Dear Arthur
>
> The problem was fixed with the new version of ASREML in the
> following case:
>
> BLUP analysis for pilodyn and diameter - F229 F228
> tree !P
> mp
> fp 50 !A
> site 2 !I
> cross 1 !I
> pop 15 !A
> fam 50 !I
> blo 8 !I
> mplot 120 !I
> parc 400 !I
> code 30 !A
> pil
> diam
> .....F228_9.sas !ALPHA !MAKE
> .....F228_9.sas !MAXIT 1
> pil diam ~ Trait.site Tr.site.blo !r Tr.tree,
> -at(site,2).mplot Tr|1/at(site,2).mplot 0.3 !GF,
> -at(site,2).parc Tr|1/at(site,2).parc 0.4 !GF
> 2 2 1 !STEP .1
> 800
> 2 0 US 0.000001 0.000001 535 !GFFF
> 1600
> 2 0 US 3.9 20.1 348 !GFFF
> Tr.tree 2
> 2 0 US 1.9 12.84 135.5 !GFFF
> tree
>
>
> However, the error message "Unable to invert R or G matrix", still
> appeared in the case where the order of the sites and traits within sites
> was changed in the data file. Here the as.file was:
>
> BLUP analysis for pilodyn and diameter - F228 F229
> tree !P
> mp
> fp 50 !A
> site 2 !I
> cross 1 !I
> pop 15 !A
> fam 50 !I
> blo 8 !I
> mplot 120 !I
> parc 400 !I
> code 30 !A
> diam
> pil
> ....F228_9a.sas !ALPHA !MAKE
> ....F228_9a.sas !MAXIT 1
> diam pil ~ Trait.site Tr.site.blo !r Tr.tree,
> -at(site,1).mplot Tr|2/at(site,1).mplot 0.3 !GF,
> -at(site,1).parc Tr|2/at(site,1).parc 0.4 !GF
> 2 2 1 !STEP .1
> 1600
> 2 0 US 348 20.1 3.9 !GFFF
> 800
> 2 0 US 535 0.000001 0.000001 !GFFF
> Tr.tree 2
> 2 0 US 135.5 12.84 1.9 !GFFF
> tree
>
>
> As the analysis in both cases was the same, why did ASREML crash
> following the change of the order in the data?
>
> Sorry to insist with this problem, but it is important for me to understand
> why it happened, before I start to organize all the data files for analysis
> (where the situation is: one trait measured in all sites and two traits
> missing in some sites).
> I am sending the data and as.files corresponding to the case where the
> error message appeared.
>
>
> Was I correct, regarding the question of the example 8.4, "Multi site
> analysis", of the manual"?
>
>
> Thanks again for the help.
> Best regards.
>
> Joao
> (jces@kvl.dk)
>
>
>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
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 still free by anonymous ftp from pub/aar on ftp.res.bbsrc.ac.uk
or point your web browser at ftp://ftp.res.bbsrc.ac.uk/pub/aar/
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
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml
<> <> <> <> <> <> <>
"You (Father) have given Him (Your Son) authority over all flesh,
that He should give eternal life to as many as You have given Him.
And this is eternal life,
that they may know You, the only true God,
and Jesus Christ whom You have sent." John 17: 2-3.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml