Path analysis with nominal variables
Message/Author
 Andy Ross posted on Tuesday, April 14, 2009 - 7:56 am
Dear Prof Muthen

I would like to check that I am specifying my model correctly

When estimating a path analysis with a nominal variable which serves as both a dependent and independent variable - do I have to create dummy variables for the part in which I treat it as an independent?

ie:

usevariables = u1 u2 u2a u2b u2c u3 u4;
nominal = u1 u2;

Model:

u1#1 u1#2 u1#3 on u2a u2b u2c u3;
u2#1 u2#2 u2#3 on u4;

(where both u1 and u2 are 4 level nominal variables - and I have constructed 3 dummy variables u2a u2b u2c to represent u2 when treating it as a predictor)

I did also try:

u1#1 u1#2 u1#3 on u2#1 u2#2 u2#3 u3;
u2#1 u2#2 u2#3 on u4;

But I got an error message that nominal variables cannot appear on the right hand side of an on statement.

Andy
 Linda K. Muthen posted on Tuesday, April 14, 2009 - 8:54 am
A nominal variable cannot be used as a mediator. You could look at each category separately or use the nominal variable as a grouping variable.
 Andy Ross posted on Wednesday, April 15, 2009 - 5:52 am
Many thanks for your response - I don't think my original question was very clear

What I'm trying to do is estimate the equivalent of a latent transition analysis (example 8.13 without the interaction) - except that in my analysis c1 and c2 are not latent, and u3 and u4 are the same categorical predictor measured at two time points (in that sense I'm not examining u2 as a mediator of u4 on u1)

How would I then specify the regression of u1 on u2?

Many thanks

Andy
 Linda K. Muthen posted on Friday, April 17, 2009 - 11:00 am
See Slides 10-30 of the Topic 6 course handout at http://www.statmodel.com/newhandouts.shtml. In particular, Slide 13 shows how to make the observed variable u equivalent to the categorical latent variable c.