On 6 Feb 2003 05:27:37 -0800, [EMAIL PROTECTED] (bayene) wrote:

> Dear all,
> 
> I want to perform Confirmatory Factor Analysis by using AMOS 4.0.I
> have the scales of likert (22 items) and non-normal distribution of my
> data. 

If you are calling it 
"non-normal" by virtue of being Likert, 
that's not apt to be a problem.

>      Can I use spearman's correlations like input data for AMOS ? 

Using Spearman's  is equivalent to performing
a rank-transformation, then doing Pearsons.
Using the rank-transformation is not good when 
there are a lot of ties, or when the correlations are
especially high -- Don't you have both problems?

>  or polychoric 's correlations ?

Some folks like them.  I just know that for tetrachoric
r's, there is enough extra error, you need (say) twice
the N.  Do you have N=450  for your 22 items, 
instead  of N=220?   However,  I don't know whether
tetrachoric is worse or better than polychoric for N.

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to