Re: estimability

Dear juan
this is simply due to the collapsing over the classify set. As you have
both e and b in the model as fixed effects and you request as table of t x
e predictions then the full table of cells from which we collapse to get
t.e is the t x e x b table. This table has missing cells, specifically the
e2b3 cell is missing and hence the resulting predictions formed by
averaging over the missing cell will be non-estimable.

There are numerous work around solutions like using present and so on, but
I could not recommend what your best solution is without knowing the
application and the aim.. For example you may wish to predict at b=2 or
b=1 or for the average of b =1 and 2, and so on

warm regards

Brian Cullis
Research Leader, Biometrics &
Senior Principal Research Scientist
NSW Department of Primary Industries
Wagga Wagga Agricultural Institute

Professor,
Faculty of Agriculture, Food & Natural Resources
The University of Sydney

Phone: 61 2 6938 1855
Fax: 61 2 6938 1809
Mobile: 0439 448 591
Received on Mon Feb 01 2008 - 08:18:41 EST