Hi,
I am working with a model with normal errors but a log link. ASREML does
not seem to handle this for my dataset. Part of the problem may be that
the log likelihood is positive and is also relatively flat. I would like
to profile the variance components to assess this.
Can I use the log likelihood to manually search a small grid?
Previous versions of the manual had discussed the use of the BLUP and
MAXIT options. Is it correct to use MAXIT=1 or BLUP 1?
I think that the !BLUP v option is broken or doesn't work as indicated in
the recent versions of ASREML if v is not 1 (see below).
Thanks,
Bruce
Normal with log link Glmm
animal !p
perm !A
breed !A
year !I
age !I
or !M 0
nlb !M 0
ped.coded
brs.data !NODISPLAY !blup 2 # !maxit 1
or !log !disp !RESPONSE ~ mu breed year age !r animal 0.2 perm 0.09
Provides :
$ asreml z
13 Sep 2000 11:52:49.800 8.00 Mbyte Unix z
A-inverse retrieved from ainverse.bin
PEDIGREE [ped.coded ] has 854 identities, 2823 Non zero elements
Univariate analysis of or
Forming 1231 equations: 19 dense
1 -269.14 0.508 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.53
1 816.66 0.475E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.19
1 781.90 0.513E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.19
1 688.32 0.596E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.21
1 652.94 0.644E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.20
1 647.85 0.669E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.19
1 647.70 0.670E-01 921
>>>>> >>>>> >>>>> >>>>> Iteration complete U+S=T 0.19
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml