Re: ASREML crashes II
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ASREML crashes II



Dear Hermann,

There are three problems I will comment on.
The major problem is the layout of the starting values
for you variance matrices.

The syntax is not quite right.  If the initial values were all
on the  TRait 0 US  line, it would be correct.  The way to
split them over several lines is to use the !+ qualifier.

For example you had 

Tr.arbol 2
Trait 0 US 0.0652 0.15 0.0708 0.15 0.15 9.995 0 0 0 0.0618 0 0 0 0.15 0.0461,
0 0 0 0.15 0.154.3553 !GP 
arbol

but you need to have

Tr.arbol 2
Trait 0 US !+21 !GP
0.0652 0.15 0.0708 0.15 0.15 9.995 0 0 0 0.0618 0 0 0 0.15 0.0461,
0 0 0 0.15 0.154.3553  
arbol

NB The !GP qualifier must be on the Trait 0 US line

The program was crashing because it could not find enough values on the line.
[ I have inserted a check so in future it will die more graciously I hope.]
  So if you change the job to 

Prueba de evaluacion de ensayos de E globulus de Mel-Mel
 arbol !P
 familia 199
 localidad 2
 bloque 5
 procedencia 26
 altura1  !M 0
 dap1     !M 0
 v1       !M 0
 altura2  !M 0
 dap2     !M 0
 v2       !M 0
PEDIGREE.TXT !alpha !make
MULTIVAR.TXT !skip 1 !maxit 10 #!filter 3 !select 1
altura1 dap1 v1 altura2 dap2 v2 ~ Trait !r Tr.arbol Tr.bloque Tr.procedencia,
Tr.bloque.procedencia Tr.bloque.familia

1 2 5 # !STEP .01
0
Trait 0 US !GP !+21
0.289 0.15 0.0864 
0.15 0.15 44.0711
 0 0 0 0.2252
 0 0 0 0.15 0.0674,
0 0 0 0.15 0.15 18.9119 

Tr.arbol 2
Trait 0 US !+21 !GP
0.0652 0.15 0.0708 0.15 0.15 9.995 0 0 0 0.0618 0 0 0 0.15 0.0461,
0 0 0 0.15 0.154.3553  
arbol

Tr.bloque 2
Trait 0 US !+21 !GP
0.0122 0.15 0.0010 0.15 0.15 0.6296 0 0 0 0.0036 0 0 0 0.15 0.0016,
0 0 0 0.15 0.15 0.2917 
bloque

Tr.procedencia 2
Trait 0 US !+21 !GP
0.0840 0.15 0.0263 0.15 0.15 7.1093 0 0 0 0.0528 0 0 0 0.15 0.0141,
0 0 0 0.15 0.15 2.6531 
procedencia

Tr.bloque.procedencia 2
Trait 0 US !+21 !GP
0.0777 0.15 0.0225 0.15 0.15 10.4676 0 0 0 0.0655 0 0 0 0.15 0.0206,
0 0 0 0.15 0.15 3.7395 
bloque.procedenci

Tr.bloque.familia 2
Trait 0 US !+21 !GP
0.0612 0.15 0.0173 0.15 0.15 10.7501 0 0 0 0.0221 0 0 0 0.15 0.0051,
0 0 0 0.15 0.15 1.3344 
bloque.familia

It will at least read the parameter values.



The second rather subtle problem is that it failed
to read the 
Tr.bloque.procedencia 2
Trait 0 US !+21 !GP
0.0777 0.15 0.0225 0.15 0.15 10.4676 0 0 0 0.0655 0 0 0 0.15 0.0206,
0 0 0 0.15 0.15 3.7395 
bloque.procedenci

structure [ notice I have changed it by deleting the 'a' from the end.
The explanation is that Tr.bloque.procedencia is 21 characters but
only 20 characters are retained.  So it was truncated to Tr.bloque.procedenci
although this s not evident in the output because I infact only print about
12 of the 20 characters in the factor summary.  Anyway, ASREML wants to form
Tr.bloque.procedenci  which it does by forming an interaction of
Tr (which it recognises as Trait) and bloque.procedenci.  It then
defines bloque.procedenci as the interaction of bloque and 
(which it recognises as procedencia).

Thats al OK until it comes to defining structures and goes looking for
bloque.procedencia
which is not recognised as being
bloque.procedenci

So I needed to drop the final a from bloque.procedencia
so it would recognise the structure definition.

Finally,
Now we can read the job in but you have 6  6x6 variance structures to
estimate and this is likely to much for ASREML to sort out
just diving in so you will need to procede more slowly.

I suggest you use the !CONTINUE qualifier and initially work on
one matrix at a time.  i.e.  change  !GP to !GF for all structures 
except the one being estimated this time.  So it will takes six runs 
to update each of the matrices once.

When ASREML runs, it writes the current parameters to multivar.rsv

The !CONTINUE qualifier reads that file to get the latest parameter values
before continuing the iteration.  This will take a while so I'm
not about to do it for you but wish you luck.

Arthur

<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD                  mailto:Arthur.Gilmour@agric.nsw.gov.au
Principal 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
mailto:asreml-request@chiswick.anprod.CSIRO.au

To send messages to the list, mailto:asreml@chiswick.anprod.CSIRO.au

Asreml list archive: http://www.chiswick.anprod.csiro.au/lists/asreml

                        <> <> <> <> <> <> <>
"Christ Jesus came into the world to save sinners"   I Timothy 1:15.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>






--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml