A query for users of PROC NESTED:

We would like to know if anyone else has used nested ANOVA where the hierarchy of nesting was somewhat ambiguous, or simply has experimented with alternate sequences of variables in the CLASS statement. We have an unbalanced data set, in which the structure of data collection does not dictate a clear hierarchy such as that in the example given in the SAS Manual. We also have a strong 'a priori' belief about how the nesting should be structured, based on which variables ought to determine other variables. But as we were trying to understand better how the procedure works, degrees of freedom, etc., we tried reversing two of the terms in the nesting hierarchy. It had such a dramatic effect on our results that we were left a bit uncertain as to how to report them. Clearly our findings are correct only if our 'a priori' hypothesis is also correct - but there is no obvious way to test the latter.

Has anyone else looked at this?

Also, who can suggest a more up-to-date biostat text on nested ANOVA than the refs in the SAS Manual?

Thanks a lot in advance.

shy