Mplus Discussion >> SEM with a nominal dependent variable

Topics
Last Day
Last 3 Days
Last Week
Tree View

Edit Profile


SEM with a nominal dependent variable

Mplus Discussion > Structural Equation Modeling >

Message/Author

Katayoun Safi posted on Wednesday, October 28, 2009 - 2:15 am

Hi,
how can I estimate a SEM with a nominal dependent manifest variable? If I include the command "nominal is", the program keeps saying that my estimation is only possible with monte carlo. Isn't there a more simple solution for my problem? Thank you for any help.

Linda K. Muthen posted on Wednesday, October 28, 2009 - 6:30 am

The message refers to Monte Carlo integration. You need to add INTEGRATION=MONTECARLO; to the ANALYSIS command.

Katayoun Safi posted on Wednesday, October 28, 2009 - 7:27 am

Thank you! Is there a way to choose the reference category?

Linda K. Muthen posted on Wednesday, October 28, 2009 - 7:48 am

Not without reordering the variable. Mplus uses the last category as the reference category.

Katayoun Safi posted on Thursday, October 29, 2009 - 1:14 am

Many thanks. Could you recommend a paper that addresses the topic of a SEM with a nominal dependent observable variable?

Linda K. Muthen posted on Thursday, October 29, 2009 - 7:09 am

I don't know of any offhand.

Katayoun Safi posted on Friday, October 30, 2009 - 5:38 am

Thanks anyway. Is there a way to receive other output than odds ratios in my SEM with a nominal dependent variable, for example maximal effects?

Linda K. Muthen posted on Friday, October 30, 2009 - 6:09 am

No. There are no extra options in this case.

sandra C. posted on Wednesday, April 28, 2010 - 4:44 am

Hello!
We are estimating a SEM with nominal dependent variable (3 categories). Could you recomend us a way to analyse the godness of fit of the model, please?
Thanks in advance.

Linda K. Muthen posted on Wednesday, April 28, 2010 - 5:11 am

When chi-square and related fit statistics are not available, nested models are compared using -2 times the loglikelihood difference which is distributed as chi-square.

sandra C. posted on Wednesday, April 28, 2010 - 8:53 am

Thanks, Linda. Do you know if, there is any way to perform multigroup analysis with nominal dependent variable?

Linda K. Muthen posted on Wednesday, April 28, 2010 - 10:22 am

You would need to do it with TYPE=MIXTURE and the KNOWNCLASS option.

Hicham Ra�q posted on Monday, October 29, 2012 - 1:24 pm

Im working on SEM with nominal dependent variable (4 categories). What il the main important criterion to test model goodness of fit. And how could we interpret it.

Thanks

Linda K. Muthen posted on Tuesday, October 30, 2012 - 1:29 pm

There are no absolute fit statistics in thia situation. Nested models can be compared using -2 times the loglikelihood difference which is distributed as chi-square or BIC.