Re: predict statement

From: Anne Winkelman <awinkelman_at_LIC.CO.NZ>
Date: Thu Apr 6 09:06:55

Dear Arthur,

1. Would it be possible for you to supply us with version 1.99c?

2. I tried using ide as you suggested. The model and predict statements
were as follows:

fyld ~ pfrn pjer agelin agequad,
 !r animal 0.7 ide(animal) 0.8 !f cgroup mv
predict animal 1:80 !ONLYUSE animal ide(animal)

The output (.pvs) file once again contained only zeros for the estimates:

 Predicted values of fyld
 agelin is evaluated at average value of 0.0000
 pjer is evaluated at average value of 0.4821
 pfrn is evaluated at average value of 0.4977
 cgroup is averaged over fixed levels
 agequad is evaluated at average value of 0.0718
  Model term animal is included in this
  Model term ide(animal) is included in this

 animal Predicted_Value Standard_Error Ecode
        99640 *************** 0.0000 E
     12237320 0.0000 0.0000 E
       362308 0.0000 0.0000 E
       364158 0.0000 0.0000 E
       364234 0.0000 0.0000 E
       364672 0.0000 0.0000 E
       365164 0.0000 0.0000 E
       100059 0.0000 0.0000 E
      9038920 0.0000 0.0000 E
     13654223 0.0000 0.0000 E
       100203 0.0000 0.0000 E
       100243 0.0000 0.0000 E

So it appears that whether we fit 'ide(animal)' or 'cow', the predicted
estimates are always zero. Perhaps there is a bug in the software?


             Sent by: ASReml To
             users discussion ASREML-L_at_AGRIC.NSW.GOV.AU
             group cc
             SW.GOV.AU> Subject
                                       Re: predict statement

             05/04/06 04:28 PM

             Please respond to
               ASReml users
             discussion group

Dear Anne,

1. The version we are using is: ASReml 1.99a [01 Aug 2005]
It says: This Beta version is valid for 9 months from 01 Aug 2005

** You could update to 1.99c but that is not the main problem I think

2. The variables were coded as you indicated:
 animal !P # variable 1
 cow !I 6018 # variable 2

3. There are not several animals per cow simply because we don't have
repeated records for cows in this particular data set. However I am doing
the analysis with a view to extending the procedure to cases in which we
have multiple records per animal.

**I presume then that there are several records per animal.
But with a single animal per cow, fit ide(animal) rather than cow.
It should have the same variance component, but then you can use

predict animal 1:80 !only animal ide(animal)

4. The use of !PRESENT did not appear to improve the estimates (they are
all still zero).

**I would need to run the job here to see what is happening.

5. Using a smaller number of cows did not improve the estimates either.

The reason I wanted the standard error of prediction of the sum of the
animal and cow effects was to calculate reliabilities of the production
value, which is the sum of the animal (genetic) and cow (permanent
environmental) effects. I was hoping that the estimate would include the
diagonal elements of the two effects and off-diagonal elements associated
with the two effects from the inverse of the mme. Perhaps this is asking
too much of the software?

Yes. You really want the coefficients from the C inverse (which can be
extracted using !CINV qualifier)
but ASReml uses a sparse C inverse which probably does not include the
covariances you want.

If you could work out that say animal 3 was born to cow 1, you could
predict animal 3 0 3 cow 0 1 1 !parallel animal cow !vpv !onlyuse animal

for a particular combination but not for 6000 odd combinations.

 Check out
for my personal greetings and update.

I thank God for His mercy through Jesus Christ.

Arthur Gilmour PhD
Principal Research Scientist (Biometrics)
NSW Department of Primary Industries
Orange Agricultural Institute, Forest Rd, ORANGE, 2800, AUSTRALIA

fax: 02 6391 3899; 02 6391 3922 Australia +61
telephone work: 02 6391 3815; home: 02 6364 3288; mobile: 0438 251 426
The ASReml discussion group is at
To join it,

Proposed travel:
ASReml workshop CSIRO 16-18 May
[IBC Quebec 16-21 July 2006]
WCGALP8 13-18 Aug Brazil

This message is intended for the addressee named and may contain
confidential information. If you are not the intended recipient or
received it in error, please delete the message and notify sender. Views
expressed are those of the individual sender and are not necessarily the
views of their organisation.
Received on Tue Apr 06 2006 - 09:06:55 EST

This webpage is part of the ASReml-l discussion list archives 2004-2010. More information on ASReml can be found at the VSN website. This discussion list is now deprecated - please use the VSN forum for discussion on ASReml. (These online archives were generated using the hypermail package.)