Mike Wogan wrote:
> Rich,
>
> You've gone way beyond me with your question.
>
> I suggest you try posting your question to a statistics list. Here's
> one that I used to belong to. These are all people teaching statistics,
> working at SPSS, etc., and someone on the list will be able to help you:
>
> [EMAIL PROTECTED]
>
> that's edstat-L (not "one").
>
> Drake Bradley, at Bates, has written a program for Experimental Psych
> which will generate *data* randomly, which conforms to certain limits
> (mean, s.d.), but I'm not sure his programs will do anything that complex.
>
> Mike
Datasim generates correlated datasets as well, both for multivariate designs
and for correlated groups experimental designs (repeated measures,
split-plot, etc.). However, it requires that the population correlation
matrix be one that can actually arise in practice, and has no procedure for
generating a matrix as close to the user's (ill-defined) target matrix as
possible.
Drake Bradley
