Predicting nesting values....

All,
 
I have some yields (tch) from a set of sugar trials done at 4 locations
(Loc) and 2-3 years (Cclass). The 4 locations have been nested into two
groups x two locations with LocGroup as a measure of root pathogens (1 =
clean, 2 = infected).
 
I'd like to compare the genotype (Gen) BLUPs in either clean or infested
trials. However, the predict statements below, average the mean yield
over both clean and infested trials for either the trial (predict
LocGroup) or genotype means (predict Gen LocGroup).
 
How can I modify the model to predict at the marginal effects of
LocGroup, given that I will not get a prediction if it is included in
the fixed part due to aliasing with Env. Can I only do this if I
construct a new factor that is Env within LocGroup and then use LocGroup
and the new factor in the model?
 
sugar stuff...
 Reg !A 1
 Env !A 9
 Loc !A 4
 Cclass !A 3
 Column !I 6
 Row !I 24
 Rep !I 2
 Gen !A 51
 tch
 LocGroup!I 2
 LocPath !I 4
n.csv !skip 1 !mvinclude !MAXITER 100 !DOPART 3

!PART 3
tch ~ mu Env mv !r Gen Gen.Cclass Gen.LocGroup Gen.LocGroup.Env
9 2 0
4 Column AR 0.2 !S2=249
24 Row AR 0.4
etc
 
predict LocGroup
predict Gen LocGroup
 
 

Dr. Scott Chapman

CSIRO Plant Industry, ph/fax +61 (0)7 3214 2254/2950

 <mailto:scott.chapman_at_csiro.au> scott.chapman_at_csiro.au

 
Received on Tue Nov 02 2004 - 18:10:19 EST