-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
Behalf Of John Uebersax
Sent: Thursday, June 27, 2002 4:08 PM
To: [EMAIL PROTECTED]
Subject: Re: SEM and measures of fit.
[EMAIL PROTECTED] (nothanks) wrote in message
news:<[EMAIL PROTECTED]>...
> I ...am curious why SEM
> practitioners seem to assess model fit by only comparing the
> observed and predicted covariance matrices. As opposed to (also)
> using statistics based on observed vs. predicted outcomes.
A closely related question is why estimate an SEM from covariance
matrices instead of raw data. The former would appear to use only
some of the information (as in limited information maximum likelihood
or LIML estimation), whereas the latter would be a full-information
(as in full-information maximum likelihood or FIML estimation)
approach.
---------------------------------------------------------------------
I have thought a lot about this problem, and have not got to any good
answers yet.
SEM is basically a covariance model. To take the resulting sigma matrix and
develop a set of equations that can be used to make direct predications of
the raw data is a major problem. One has to make a lot of assumptions on
values of required parameters, since the prediction model has more
coefficients than the covariance model.
Also the focus is on variations in data about a mean value (Large data sets,
statistically give firm means), not on absolute values. The information
content of the means is not relevant, and is not part of the SEM model.
I find this unsettling.
David A. Heiser
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
