repeated measures help; disagreement with SPSS

Peter Dalgaard BSA <p.dalgaard at biostat.ku.dk> writes:

Argh. You have no idea how difficult it is to read that stuff when
you're not on a Windows machine... But I suppose that converting it to
plain text is a pain even *on* Windows.

My apologies.  I hate the program and environment too, which is one of
the reasons I'm trying to change ;-)

Anyways, as far as I can see, you are in fact getting the same
interaction test (RL*COND, Sphericity Assumed, Type III SS=17.734), so
I'd suspect that the test for the main effect is one of those weird
things where  you take the average over the three levels of cond,
ignoring the fact that one level occurs twice as often as the others.

What happens if you run the SPSS analysis without the interaction term? 

YES!  You were exactly right.  I tried to change the model in SPSS,
but it wouldn't let me in that case (it always includes the
interaction, grumble).  so I made every condition have the same
number of data points.  and that showed up as the exact same set of
numbers in SPSS and R (whew!):  same p values, same SS, MS, etc.

OK, so the next question is, I understand your reasoning about the
weird thing, but which one *should* I use, the weighted version (like
R) or the unweighted (like SPSS)?  (realistically, this probably
doesn't happen too much, and certainly not to this degree, so it
probably doesn't have much of a profound effect in any case).  This is
probably not the right place for this question, but ...

thanks a ton for your help!

greg
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._