🚀 go-pugleaf

I would like to solicit suggestions about how to overcome some
limitations with structural equation modeling (SEM) software.  We need
to run models with a large number of groups, a large number of
subjects, and in a short amount of time.

We would like to use SEM software to test a large latent growth
structural model (LGM) across several different groups.  In addition,
we wish to employ methods to account for data missing not completely
at random to reduced the bias cause from listwise deletion of cases.

To run these analyses, we need our SEM software to allow for many
different groups, like 20 or more.  We currently run EQS on our UNIX
systems.  Although EQS runs very quickly, the programmers limited the
number of groups to nine (although 10 will work).  Here lies our
problem:  we need a SEM package that both runs quickly and can work
with large data sets and matrices.  As I mentioned, we intend to use
as many as 20 groups (or more), and we have about 13,000 cases in our
sample.

We would greatly appreciate any suggestions that can get around these
limitations.  Ideally (dreaming?) we would like a modified version of
EQS that allows a greater number of groups.  Short of that, we're not
sure where to turn.

We considered Lisrel, but I believe no UNIX version has been written.
Lisrel on our VAX (near death) cannot handle the jobs--we estimate it
will take about 12 to 15 hours for each run.  This would be far too
slow given the amount of tweaking needed to get correct start values,
and so on.

I understand that we can purchase both EQS and Lisrel for the PC.
Does the PC version of EQS allow for more than nine groups?  Even so, I
wonder how well a PC could handle both the amount of data we intend to
use and the size of the matrices required for the large number of
groups.  EQS typically uses about 20 megs of temporary space on our
UNIX boxes for some of our "small" jobs.

Thanks in advance for any help.  If this isn't clear, please e-mail me
for more info.  Oh, and we would be willing to consider other SEM
software too.

-keith

 ______________________________________________________________________
                                                              _    _
 o r e g o n  r e s e a r c h  i n s t i t u t e             ( )  |<  |
 _______________________________________________

 e u g e n e  o r e g o n
 ________________________
                                        keith smolkowski  keith@ori.org