Re: do two SEM models differ significantly?

On 12 Aug 2006 16:50:21 -0700, "Ray Koopman" <koopman@xxxxxx> wrote:

Anon. wrote:
Ray Koopman wrote:
Burger wrote:
Suppose you have two structural equation models and you want to test wether
they differ significantly.

If I understand it correctly, you can do this by looking at the difference
in Chi-square between the two models with as the degrees of freedom also the
difference in df between the two models.

Can you also test for the significance of the difference between the two
models with an F-test? If yes, how and when is this preferred over the above
method?

The usual large-sample test for the difference between nested SEMs
treats twice the log of the ratio of the maximized likelihoods as
having a chi-square distribution. There may be adjustments for small
samples sizes and/or nonnormality that use the F distribution.

This is curious to me: how many people use chi-squared tests in an ANOVA?

SEM = Structural Equation Model, which is not a regression model. It
models a covariance matrix -- e.g., Sigma = A B A' + C, with various
constraints on A, B, and C -- and then fits it to a sample covariance
matrix S.

Ray, is there an online tutorial on SEM that you'd recommend? Or is
there a standard text?
-*** Startz
.