View topic - the pol(day,-3) in three different animal models | Forum

Author

Message

< ASReml ~ the pol(day,-3) in three different animal models

liubaosuo

Posted: Thu Mar 01, 2012 2:05 am

Joined: 12 Mar 2010 Posts: 2

Dear Professor Gilmour,
I have a problem about the pol(day,-3) in three different animal models.
Firstly, I want to fit an univariate animal model, there is one record per animal
Asreml analysis harvest body weight
Animal !P
tank !A
family !A
day !A # the covariate day(age) at harvest
bw
pedigree.dat !skip 1
bw.dat !skip 1 !MAXIT 100 !EXTRA 5
bw ~ mu tank pol(day,-3) !r animal family

Secondly, I want to fit a fixed regression animal model and random regression animal model , there are many records per animal in different test days.
Asreml analysis test day body weight
Animal !P
tank !A
family !A
day !A # the covariate test day
bw
pedigree.dat !skip 1
bw.dat !skip 1 !MAXIT 100 !EXTRA 5
#a fix regression animal model
bw ~ mu tank pol(day,-3) !r animal family

#a random regression animal model
bw ~ mu tank pol(day,-3) !r pol(day,3).animal pol(day,3).family

Are three models wrote by me right? Are the pol(day,-3) in three models the same? Are there some difference between the three models?
Thank you in advance!
baosuo

Arthur

Posted: Sat Mar 03, 2012 12:49 am

Joined: 05 Aug 2008 Posts: 277 Location: Orange, NSW

Dear bausou.

Comment 1.
day needs to be a variable, that is a number representing days since a particular date.
But you have declared it an Alpha factor, which means it is internally the numbers
1, 2, ... the number of distict values in the field. Now this might be sensible if the data
is sorted by days and measurements are taken on consecutive days. I would
expect you just need to drop !A assuming the value is 'age at harvest in days'.

Comment 2.
You have not explained the relationship between 'animal' and 'family'. Your model
suggests to me that 'animal' will pick up additive genetic effects (GCA) and family
is picking up specific combining ability. Just wondering.

Comment 3.
Model 1 specifies a cubic regression. I suppose animals are harvested at varying ages
and the response here is weight at harvest. To separate age from genetic effects,
families need to be represented across several ages.

I would usually use a spline to model the mean trend in this and the subsequent models.

Comment 4,
Model 2 is the same as model 1, but you now say the data is different, having many records per animal.
In this repeated measures context, I would also fit ide(animal) in the model
as the between animal residual. Otherwise, the genetic animal term will be inflated.
yOUR REFERENCE TO 'TANK' Suggests THESE MAY BE FISH. If so, and if there is a
big range in weights (e.g. 10 fold between first and last measurment, then
I would be concerned about variance hetergeneity. Maybe analyse log(bw)
so that animal effects relate to relative size. In anycase, in this model, the
pol() term is picking up trends in ages within animals, a different concept to
model 1.

Comment 5.
This is a random regression model. You need to specify a variance structure for the two
pol(day,3) terms. I would start by making these terms pol(day,1), and if that fits OK
try pol(day,2) before trying pol(day,3).

_________________
Arthur Gilmour

Retired Principal Research Scientist (Biometrics)

liubaosuo

Posted: Mon Mar 05, 2012 12:12 pm

Joined: 12 Mar 2010 Posts: 2

Thank you very much! i have knew what you said, i should study the Asreml user guide carefully!
Thank you again!
baosuo liu

All times are GMT

View next topic View previous topic Page 1 of 1 Forum \| VSN International Forum Index ~ ASReml

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum

Template by Mike Lothar // Protected by Anti-Spam ACP

| | | | | | |