Hi,
45 manfiest variables is no small number.
What kind of item responses will be used
(e.g., binary, ordinal)? How symetrical
to you expect the empirical item distributions
to be? Do you plan to fit your model to
covariance matrix or a matrix of tetra/poly-
choric correlations? What estimation method
do you propose to use? If you plan to use
WLS or ADF, then you will need a very large
sample (though with this many items, you may
have to give up on WLS and ADF). There are
altenatives requiring smaller samples such
as the Bollen-Stine boostrap (AMOS), Satorra-
Bentler corrections to the ML chi-square and
standard errors (EQS and LISREL), and some new
estimation methods in Mplus program.
Generally I recommend an absolute minimum of
300 per sub-sample of interest. More is better,
and is probably worth the effort if you can
recruit.
HTH
Steve
>Hello.
>
>Can someone please check my calculations before I go crazy recruiting
>well over a thousand subjects? I am trying to validate a questionnaire
>which is hypothesized to have three subscales, each measured by
>fifteen questions, using confirmatory factor analysis (structural
>equation modeling).
>
>Thanks,
>Marianne
--
----------------------------------------------
Steve Gregorich
University of California, San Francisco
Center for AIDS Prevention Studies
Medical Effectiveness Research Center
Center for Aging in Diverse Communities
74 New Montgomery Street
San Francisco CA 94105
[EMAIL PROTECTED]
http://sites.netscape.net/gregorich/index.html
----------------------------------------------
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
