Hi Peter, thanks.
After reducing the data to seven factors, these go into a regression
model to predict a continuous outcome.
I'm not familiar with proc pls -- any info on that?
Funny that you should mention MI and MIANALYZE--I'm also investigating
that for the missing data issue within this dataset.
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Sent: Thursday, May 10, 2007 11:51 AM
Subject: Re: variable reduction (proc factor) with repeated measures
Julia Rushing <jrushing@BELLOMYRESEARCH.COM> wrote
I'm working on a study with multiple variables that we are running
through PROC FACTOR (method=princomp) to reduce the highly correlated
variables down to a subset of factors (21 vars were reduced to 7
The respondents in the dataset have from 1 to 5 records depending on how
many products they rated. Responses within respondent should be
correlated. So far we've done the factor method ignoring this issue. I
wonder what techniques are available to take this within-person
correlation into account while or before doing the factor analysis.
Perhaps results would not differ.
1) What is the purpose of the data reduction? Is it for some kind of
regression model? If so, what kind?
2) Have you considered PROC PLS instead of reducing the data this way?
3) If you decide that reducing the data this way is the right thing to
do, PROC MI and PROC MIANALYZE may be helpful
and I hope this was helpful....but if you provide more context, more
helpful help may happen (please respond to SAS-L, not just me)