Rich Ulrich wrote:
> You are rather defeating the nature of the "repeated measures"
> if you stick in some measure that is different.
> Or some measure that is on a different scale.
> Or some measure that has been 'normalized'
> differently  from the others.
> 
> You might be able to do it, if you are simply wiping out
> all the between-period differences by the normalizations;
> but I'm not sure why you would do it.  Do you have a
> model that explains why you could have arbitrary and
> different treatments of periods?

I think you'd need to model the variances and the covariances of the data in
more detail - some kind of multilevel model might be suitable, though it
wouldn't be straight-forward.

Of course, a transformation might still be sensible if the the reason for the
discrepancy in the scale of one was the high mean and/or larger range relative
to the others.

> And I *might*  hesitate to call the resulting thing,
> "repeated measures ANOVA"  -- I think you have
> to apply MANOVA  tests, instead.

I'm not sure that would solve the problem - except in as much as the
sphericity asssumption might be violated in the example (which now that I
think on it is probably highly likely in the example given and may be what you meant!).

Thom
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to