thanks Sabri

AMOS and SPSS behind and IBM's data modeler power should always be considered.

thanks Sabri

Hi Qais,

It depends on the nature of my data (or your research problem):

when i have multivariate normality distribution i use ML or ULS measurement models (LISREL - the first SEM software).

Otherwise when

1. non normality (also we can use Assymptotic Distribution Free models -

WLS, for exemple - in the LISREL, but we need a big number of respondents);

2. Formative models;

3. Complexy models (many variables and many constructs)

4. Models still in development or with "soft" theory.

5. Few respondents

I use PLS (It is very "plastic")

i believe strongly that we should not have a prefer technique, but the problem (and the nature data) must conduct the decisions.

Hope this hope

Best regards

Dirceu

Thank you prof Dirceu for providing me valued information.