Message/Author
 Katherine Yu posted on Wednesday, May 04, 2011 - 10:48 am
I'm new to Mplus. Hope I'm nor repeating others questions (I tried, but didn't find it).

I'm trying to add gender as a exogenous variable. Here are my questions:

1). Can I still use ML method? Or is there any better estimation method?

2). How to control the correlation between exogenous variables? For example, I want to keep gender and objective stress uncorrelated.

3). Also for the variance of exogenous variables. For dechotomous exogenous variable, how to deal with the variance? For example, if 0=male, 1=female, variance of gender is 0.3, it's meaningless...

Thank you very much!
 Linda K. Muthen posted on Wednesday, May 04, 2011 - 11:23 am
The estimation method is determined by the scale of the endogenous variables. The model is estimated conditioned on the observed exogenous variables. Their means, variances, and covariances are not parameters in the model. When you bring these parameters into the model, you make distributional assumptions about them and treat them as endogenous variables.
 Katherine Yu posted on Wednesday, May 04, 2011 - 12:07 pm
Thanks for your response. Can you suggest me some books or articles about it? Or can I find it in the user's guide? I don't quite understand.

Here is my syntax. Is there something wrong? I did not make any distributional assumptions (or maybe I did, I don't know).

TITLE: FETAL OUTCOME;
DATA: FILE IS b.txt;
TYPE = COVARIANCE;
NOBSERVATIONS = 200;
VARIABLE: NAMES ARE BW BL HC GA SEX FLOOD IESR MEDRISK DELICOMP;

ANALYSIS:
TYPE IS GENERAL;

MODEL: GA ON DELICOMP MEDRISK ;
BW ON MEDRISK GA SEX;
BL ON GA MEDRISK FLOOD IESR SEX;
HC ON GA MEDRISK SEX;

OUTPUT:
residual stdyx;

data is

0.301
1.176 8.549
0.802 3.326 3.386
0.570 2.711 1.950 2.857
-0.030 -0.268 -0.099 0.011 0.248
0.363 4.934 -0.213 -0.529 -0.015 128.930
0.020 0.227 -0.026 0.048 0.016 2.277 0.214
-0.103 -0.806 -0.485 -0.618 0.075 2.451 0.191 2.165
-0.008 -0.189 0.000 -0.182 0.008 0.791 0.015 0.327 0.605
 Linda K. Muthen posted on Wednesday, May 04, 2011 - 12:14 pm
If you don't mention the means, variances, or covariances of the observed exogenous variables in the MODEL command, you make no distributional assumptions about them. And you don't do that. This is in line with regression analysis.