David Airey wrote (excerpted):
At the back of Neter et al., 1996 (Applied Linear Models), there is a
section on figuring out appropriate tests for ANOVA terms for any model
--------------------------------------------------------------------------------
Why wouldn't it just be -anova lntime case modality reader-? The two cross-
classified random effects are assumed to be orthogonal, so there wouldn't be
any interaction term for them, and the random effects are assumed to be
orthogonal to the fixed effect, so there wouldn't be any interaction terms
there, either. It's a cross-over study design with randomized sequence (see
documentation for Stata's -pk- suite for mention of carry-over, sequence and
period) and not a repeated measures design, so there wouldn't be any
-repeated()- option, to boot.
Setting the mean squares for case and reader equal to their expected mean
squares and solving for their sigmas will give you the variances for the two
crossed random effects. (Is it a listing of expected mean squares that's shown
in the back of _Applied Linear Statistical Models_?)
Couple of corrections to my post: I mispelled Littell, and the example of the
crossed random effects model in the help file already has a fixed effect--the
population mean (intercept).
Joseph Coveney
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
