Re: ASREML output vs SAS output

Dear Emily,

Following on from Brian's comments, there are two changes required in the coding to get the analyses to agree.

First, Rep was not flagged as a FACTOR.
Since it had 2 levels, it did the right analysis but the scale was wrong.
So changing Rep
to Rep *
fixes that.

Second. Unlike many other packages, ASReml expect you to formally
include the intercept in the model if you want it. So, you had 20 clones
and 20 df for Clone from ASReml, but 19 df from SAS.
Changing the model to
Weight ~ mu Clone !r Rep
fixes thay.

So to conclude,
running
RCB with one fixed effect
 Clone !A
 Row 5
 Plot 8
 Rep *
 Weight
 CCS
 Notes
rcb.csv !skip 1 !mvinclude !maxiter 40 !nodisplay
Weight ~ mu Clone !r Rep
0 0 0

gives results
   8 LogL=-103.911 S2= 5379.6 20 df 0.3798 1.000
 Final parameter values 0.37977 1.0000

          - - - Results from analysis of Weight - - -

          Approximate stratum variance decomposition
 Stratum Degrees-Freedom Variance Component Coefficients
 Rep 1.00 46240.0 20.0 1.0
 Residual Variance 19.00 5379.58 0.0 1.0

 Source Model terms Gamma Component Comp/SE % C
 Rep 2 2 0.379773 2043.02 0.62 0 P
 Variance 40 20 1.00000 5379.58 3.08 0 P

                                   Wald F statistics
     Source of Variation NumDF DenDF F-inc P-inc
   8 mu 1 1.0 83.45 0.069
   1 Clone 19 19.0 1.19 0.353

which agrees with SAS.

------------------------
Arthur Gilmour

Retired Principal Research Scientist (Biometrics)

-------------------- m2f --------------------

Sent using Mail2Forum (http://www.mail2forum.com).

Read this topic online here:
http://www.vsni.co.uk/forum/viewtopic.php?p=853#853

-------------------- m2f --------------------
Received on Sat Sep 18 2009 - 05:27:37 EST