In article <[EMAIL PROTECTED]>,
castlemaster <[EMAIL PROTECTED]> wrote:
> there is a test of model with chi-sq value, d.f. and p-value provided
> for the polyserial correlation coefficients in prelis 2.
> is it a test of close fit or a test of significance?
It is a significance test. The test is a test of model fit (i.e., a
test of the assumptions of latent bivariate normality, fixed,
discretizing thresholds for the 'ordinal' variable, etc.) A significant
p value implies that the model by which the polyserial r was
calculated does not apply for the data.
To the right of this is a RMSEA test of close fit. For an example, see:
http://www.ssicentral.com/lisrel/pex1.htm
> does a test of significance exist?
(see above). I don't know the basis for the test. However my guess is
that the chi-squared value is a likelihood-ratio chi-squared (G^2)
rather than a Pearson chi-squared (X^2)--at least that is so for
polychoric correlations with PRELIS.
> any reference?
If you can't locate a PRELIS manual, you might post your question to the
SEMNET discussion list.
HTH.
--
John Uebersax
[EMAIL PROTECTED]
Sent via Deja.com http://www.deja.com/
Before you buy.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
