Hi, I'm running a model in ASRemlR and the conditional wald statistics table
is producing groupings for the marginality of some variables (A's B's
and
C's in the tables below) that we are struggling to understand. From our
understanding of the model we are fitting and the reference manual, we were
expecting main effects to be grouped together (e.g Maring "A") and the
two
way interactions to be grouped together (e.g. Margin "B") and to be
marginal
with respect to main effects and then the three-way interactions to be
grouped together (e.g. Margin "C") and to be marginal with respect to
two-way interactions and main effects. However, this doesn't seem to be
happening. For example, the main effect of yr (year - of which there are 2
and is a factor) appears to be given Margin "C",which groups it with a
number of 2-way interactions and the three-way interactions in terms of
marginality. Does anyone have any suggestions as to why is this the case?
I have been told that ASReml has a general algorithm for determining the
groupings which is not fool proof. Stand-alone ASReml allows you to
redefine the groupings. What we would like to know is whether those
capabilities exist for the R version, and if not, are there other ways we
can set up the model to get the terms in the appropriate grouping.
Thanks so much for any suggestions,
Sue
> W1<-wald.asreml(m1.asreml, ssType="conditional")
asreml 3.0 (15 April 2013), Library: 3.0hj (15 November 2011), X86_64
LogLik S2 DF wall cpu
-50111.3907 0.0010 51482 09:02:03 15.9
> W1$Wald$p<-round(1-pf(W1$Wald[,2], W1$Wald[,1], 90),3)
> W1
$Wald
Df F.inc
F.con Margin p
(Intercept) 1 1772.0000 585100.0
0.000
yr 1 7.6790 9030.0
C 0.007
months 5 9.1350 10790.0
A 0.000
dayc 1 21.9900 19780.0
B 0.000
haschicks 1 38.0200 50740.0
B 0.000
chickagec 1 6.5720 14020.0
B 0.012
bsc 1 191.3000 192600.0
B 0.000
sex 1 0.7405 742.4
A 0.392
winspc 1 45.7700 44840.0
A 0.000
sinwinc 1 3.4120 2586.0
A 0.068
rainc 1 16.4300 15910.0
A 0.000
months:dayc 5 15.2400 14320.0
C 0.000
months:sex 5 5.3200 669.1
B 0.000
haschicks:sex 1 7.7330 8906.0
C 0.007
chickagec:sex 1 0.2351 288.6
C 0.629
bsc:sex 1 1.9470 1726.0
C 0.166
sex:winspc 1 23.1700 19900.0
B 0.000
sex:sinwinc 1 12.4700 5766.0
B 0.001
sex:rainc 1 0.7769 625.3
B 0.380
months:sex:winspc 10 2.7480 2258.0 C
0.005
months:sex:sinwinc 10 2.6850 1834.0 C
0.006
months:sex:rainc 10 1.7060 1706.0 C
0.091
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Sue Lewis
NERC fellow
Institute of Evolutionary Biology
School of Biological Sciences
University of Edinburgh
Edinburgh, EH9 3JT, UK
Email: sue.lewis at ed.ac.uk
Tel: +44 (0)131 6505444
http://lewis.bio.ed.ac.uk <http://lewis.bio.ed.ac.uk/>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20131029/38802e1a/attachment.pl>