Repeated measurements in one trait

Hi all, does this make sense?

My "problem" is repeated measures of one trait only.
After some experimenting, and after reading some recent communication on the
ASReml mailing list about ide(), I need somebody to possibly verify that I
am on the right way.

My data is organized like this and I want to make a repeated-measure analysis
as regards ring-density joined by tree size traits in a three-trait analysis:

 TID !P # tree id
 MID !A # mother id
 FID !A # father id
 FAM !A # family id
 BLK # block
 DIA # tree diameter brh
 HGT # tree height
 YR # ring-year
 WD # ring-width
 DN # ring-density

When tree height was recorded, an increment core was taken.
Ten annual rings were measured from the same series of years for all trees.
These ten measurements are regarded as repeated measurements of the same trait.
The values vary by year, but there is no year-trend, only some random pattern.
However, the influence of specific ring formation year is significant. The
experience that density co-varies with ring-width as such may also be accounted for.

The data file content, equal for each tree in sequence:
tid mid fid dia hgt 2001 . .
tid mid fid . . 1999 wd dn
tid mid fid . . 1998 wd dn
tid mid fid . . 1997 wd dn
tid mid fid . . 1996 wd dn
tid mid fid . . 1995 wd dn
tid mid fid . . 1994 wd dn
tid mid fid . . 1993 wd dn
tid mid fid . . 1992 wd dn
tid mid fid . . 1991 wd dn
tid mid fid . . 1990 wd dn
(next tree in the same way)
(next tree in the same way)
...

The AS file continues like this:
Dseq_asr.dat !SKIP 1 !ALPHA !REPEAT !MAKE #for tree model (TID)
Dseq_asr.dat !SKIP 1 !MAXIT 99 !EXTRA 8 #this datafile has 1424 + 10*1427 desity records
DN H01 D01 ~ Trait at(Tr,1).YR.BLK at(Tr,1).WD.YR at(Tr,2).BLK at(Tr,3).BLK !r Tr.TID at(Tr,1).ide(TID)
1 2 1
0 0 IDEN
Trait 0 US !+6 !GUFUFUU
  1.5
   0 2.5
   0 3.0 7.5
Tr.TID 2
Trait 0 CORGH !+6 !GUUUUUU # optionally fix to 0 some bad correlation
 0 0 0 3.8 1 1
TID 0 AINV

Does this make sense?

I assume that the result should now reflect the three traits as such and that the
independent error term reflects the residual variation in density within trees.
The independent error term should not (?) be used in, e.g., heritability estimation.

Remains: How do I estimate residual covariation between all three traits?
That is, the within-tree covariation between density and the others?
The present set-up does not permit that, but it is probably possible?
Some anomalies in estimated parameter values make me suspect that the
model is significantly incomplete in that respect!

Best regards,

Tore Ericsson, PhD
Skogforsk, Box 3
SE-918 21 Sävar, Sweden
+46(90)2033355 (office)
+46(90)2033360 (fax)
+46(70)3198352 (mobile)
tore.ericsson_at_skogforsk.se
Received on Sun May 23 2005 - 16:10:28 EST