Author 
Message 
← ASReml ~ Models with parents and heterotic groups

Chloe 
Posted: Tue Oct 13, 2009 1:30 pm 


Joined: 23 Jul 2009
Posts: 19

Dear all,
I'm trying to fit a model with the effect of the environment (fixed), the 2 parents of the inbreeds and their heterotic groups (random).
Here is the command line:
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r GrHet1 GrHet2 and(GrHet2),
GrHet1.Parent1  GrHet2.Parent2 and(GrHet2.Parent2)
I'd like to retrieve the effet of the parents (I don't mind if they are Parent1 or Parent2). Of course, one parent corresponds to one heterotic group, so these factors are associated.
However, I don't know how to write the "predict" statment.
what is the good way to write this???
Thank you for answering!
Chloe 


Back to top 

Arthur 
Posted: Tue Oct 13, 2009 11:18 pm 


Joined: 05 Aug 2008
Posts: 471
Location: Orange, NSW

Dear Chloe,
I am struggling to understand the coding.
I understand that you have progeny from crossing parents representing 2 heterotic groups. E.g.
HG1 has parents A B C
HG2 has parents D E F
and data has columns
Parent1 Parent2
A E
B E
B F
C D
C F
...
However, you coding also has factors GRHet1 and GrHet 2
which I do not understand unless you either have moren than
2 heterotic groups and/or Parent1 and Parent2 refer to
male and female
so that the data looks like
FemP MaleP HGFem HGMale
A D 1 2
D A 2 1
B D 1 2
B E 1 2
F B 2 1
F C 2 1
...
Then HGFem and(HGMale) FemP and(MaleP)
doesn't make sense because
HGFem and(HGMale) is always 1 1
FemP and(MaleP) picks up the general combining (additive) effect
of the parents.
HGFem FemP and(MaleP)
would give n average effect for which HG was the amle parent.
So until I can understand the model better, and then what effects
need to be combined in the PREDICT statement, I can't help you. 
_________________ Arthur Gilmour
Retired Principal Research Scientist (Biometrics) 

Back to top 

Chloe 
Posted: Wed Oct 14, 2009 7:24 am 


Joined: 23 Jul 2009
Posts: 19

Dear Arthur,
Thank your for looking at my problem.
My data look like this:
Recolte Plantes Environnement Parent1 Parent2 GrHet1 GrHet2
20 100 E1 A G 1 4
30 150 E1 A F 1 3
...
This is maize data.
I'd like to analyse the number of harvested plants ("Recolte"), in relation to the number of sowed plants ("Plantes"), in different environments (I have 7 different environments), each plant being the results of the cross between two parents. Each parent belongs to one and only one heterotic group.
I have 845 parents divided in 4 different heterotic groups.
What do you advise to me?
Best regards,
Chloe 


Back to top 

Arthur 
Posted: Wed Oct 14, 2009 11:39 am 


Joined: 05 Aug 2008
Posts: 471
Location: Orange, NSW

Dear Chloe,
I would write
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r GrHet1 and(GrHet2) Parent1 and Parent2
predict Parent1 !present GrHet1
Issues:
ASReml needs to be tricked into coding GrHet and Parent conformably.
This is easier in ASReml3 where you can use
Parent1 !A 845
Parent2 !AS Parent1
GrHet1 *
GrHet2 *
Again, the predict may not work as you hope if not all 845 parents appear as
Parent1 since ASReml will not be able to work out the link based on
Parent1 and GrHet1 for those parents that do not appear as Parent1.
You may need a second predict statement
predict Parent2 !PRESENT GrHet2
That should get you started at least.
GrHet1.Parent1  GrHet2.Parent2 and(GrHet2.Parent2) 
_________________ Arthur Gilmour
Retired Principal Research Scientist (Biometrics) 

Back to top 

Chloe 
Posted: Wed Oct 14, 2009 3:43 pm 


Joined: 23 Jul 2009
Posts: 19

Dear Arthur
I tried what you suggested, that is to say:
Parent1 !A 845
Parent2 !AS Parent1
GrHet1 *
GrHet2 *
Plantes 1
Recolte 1
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r GrHet1 and(GrHet2) Parent1 and Parent2 # "and" is like "+" ????
predict Parent1 !present GrHet1
predict Parent2 !present GrHet2
I obtained a R² between predicted and observed value of 0.78.
I expected to get 2 parts in the .pvs file (because of the 2 "predict" statments), but I get only 1 part. Why?
A lot of the predicted values in the .pvs file are the same. I don't know if I should expect that.
Moreover, I compared the .pvs of this model to the .pvs of
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r Parent2 Parent1 and(Parent1)
predict Parent1
> Model M0
The results are poorly correlated.
Then I tried
Parent1 !A 845
Parent2 !AS Parent1
GrHet1 *
GrHet2 *
Plantes 1
Recolte 1
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r GrHet1 and(GrHet2) Parent1 and(Parent2)
predict Parent1 !present GrHet1
I obtained a R² between predicted and observed value of 0.71. But the results were more consistent with M0 (whose R² is 0.76).
...
Have a nice day,
Chloe 


Back to top 

Arthur 
Posted: Thu Oct 15, 2009 12:18 am 


Joined: 05 Aug 2008
Posts: 471
Location: Orange, NSW

Dear Chloe,
OOPS. Sorry.
PREDICT should be
predict Parent1 !present Parent1 GrHet1
predict Parent2 !present Parent2 GrHet2
I can't comment on the models without seeing the variance components
reported.
You should spot check the predictions by manually adding the appropriate
effects from the .sln file. Thats the only way of being sure you know
what you are getting.
On Wed, 20091014 at 16:43 +0100, Chloe wrote:
Quote: predict Parent1 !present GrHet1
predict Parent2 !present GrHet2

Best Wishes,
Arthur Gilmour
Adjunct Professor
School of Computing and Mathematics
Charles Sturt University
Jesus went to the synagogue in every town in Galilee to preach and drive
out demons.
A leper came, knelt down and said, "If you will, you can make me clean".
Jesus was moved. He reached out and touched him and said
"I am willing. Be clean." He was cleansed immediately.
Mobile Number +61 427 227 468
Home phone +61 2 6364 3288 Skype: Arthur.Gilmour
http://www.CargoVale.com.au/asreml
Travel:
Adelaide AAABG 27Sept to 2 Oct
ASReml workshop 68 Oct
Brisbane Probe 29Nov to 6 Dec
Bangladesh Prosihhkon 31Dec  31 Jan
This message is intended for the addressee named and may contain confidential information. If you are not the intended recipient, please delete it and notify the sender. Views expressed in this message are those of the individual sender, and are not necessarily the views of their organisation.
Post generated using Mail2Forum (http://www.mail2forum.com) 
_________________ Arthur Gilmour
Retired Principal Research Scientist (Biometrics) 

Back to top 

Chloe 
Posted: Thu Oct 15, 2009 3:42 pm 


Joined: 23 Jul 2009
Posts: 19

Dear Arthur,
It has corrected the problem of repeated values in the .pvs file. Thank you.
I provide you the dataset and the .as file defining the 2 models I'd like to compare.
In the .as file, I think the only first predict statment is needed.
And I wonder what is the meaning of the "and" that you added between Parent1 and Parent2.
I am still compulsing the user guide to find the better way to define the second model. I would naturally write:
"!r GrHet1 and(GrHet2) Parent1 and(Parent2)
predict Parent2 !PRESENT GrHet2"
as Parent2 and GrHet2 are the base factor for "GrHet1 and(GrHet2)" and "Parent1 and(Parent2)" respectively, but this predict statment fails.
What is your opinion?
Best regards,
Chloe 
Description: 
This is the dataset in .csv format. 

Download 
Filename: 
MaizeProd.txt 
Filesize: 
25.41 KB 
Downloaded: 
1360 Time(s) 
Description: 

Download 
Filename: 
MaizeProd.as 
Filesize: 
609 Bytes 
Downloaded: 
1986 Time(s) 


Back to top 

Chloe 
Posted: Fri Oct 23, 2009 1:02 pm 


Joined: 23 Jul 2009
Posts: 19

Dear Arthur,
The main problem I have with the following model you proposed is that I 'm not able to compute the predicted values in the .pvs neither in the .yht from the .sln .
Recolte !BINOMIAL !LOGIT !TOTAL Plantes ~ mu Environnement,
!r GrHet1 and(GrHet2) Parent1 and Parent2
predict Parent1 !PRESENT Parent1 GrHet1
What are the effects to sum to get the .pvs file from the .sln?
Thank you,
Chloe 


Back to top 

