Re: Whether parameters of structure equation model can tell the different importance of some factors

On 3 Dec 2003 04:12:23 -0800, [EMAIL PROTECTED] (Steven Wang) wrote:

> Now, i am trying to build a structure equation model based on the
> categorical data. My main aim is to find which variable, including
> gender, age, education level, is the most important factor to foster

Categorical?  I would seldom think of age that way.  Nonlinear, maybe.
Gender?  If it is  2-valued, it can be regarded as continuous.

When I try to break "education level"  into multiple categories, 
I keep approximating something that looks a lot like 
"socio-economic status" -- a different continuous measure, 
but still continuous.

Regression coefficients are easier with continuous variables, 
but I wonder why the examples of categories were so lame.


> the improvement of knowledge.

I wonder what that phrase means.  '... foster the improvement'?
> 
> anyone can give some some advice,and whether parameters of SEM can
> tell the different important of various vairables?

Regression is pretty poor at telling importance, when 
the answer is not clear-cut.   But any nested regression is 
more specific than SEM, yes?

As I understand the potential,  a test in SEM  tells you 
that some model is  particularly worse than the others.
Neither an overall test nor tests on coefficients will tell
between close competitors -- SEM is even less meaningful
that way than ordinary regression.

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."