SEM with categorical DV?

[Bianco, Joseph]

Hello:

 

I’m trying to test a mediational model in which all of the predictors (including the mediator) are continuous. The dependent variable, however, is dichotomous.  So far, I have run linear regressions to test predictors of the mediator and then I’ve used binary logistic regression for everything else. I’m concerned that testing the model in this way  (i.e., in halves, using 2 types of regression) introduces more problems than it’s worth. My understanding is that SEM requires a continuous DV, so I’m not sure what  type of analysis would permit me to analyze the model as a whole. I’d appreciate any suggestions (and thoughts on the validity/advisability of testing my model in the way I described).

Thanks so much

Joe

 

 

 

Joseph A. Bianco, Ph.D.

Department of Geriatric Medicine/Gerontology

Ohio University College of Osteopathic Medicine

353 Grosvenor

Athens, OH 45701

[phone]

[hidden email]

 

[William Dudley WNDUDLEY]

Joseph

You could look at the macros provided by Hayes and Preacher
http://www.comm.ohio-state.edu/Hayes/sobel.htm

Or Jasti and Dudley

http://www.uncg.edu/HHP/oor/

And or you could look to using a SEM package that allows binary outcomes such as MPLUS



William N. Dudley, PhD
Associate Dean for Research
The School of Health and Human Performance Office of Research
The University of North Carolina at Greensboro
126 HHP Building, PO Box 26170
Greensboro, NC 27402-6170
VOICE 336.2562475
FAX [phone]

"Bianco, Joseph" <[hidden email]>
Sent by: "SPSSX(r) Discussion" <[hidden email]>

09/02/2009 04:16 PM

Please respond to
"Bianco, Joseph" <[hidden email]>

To	[hidden email]
cc
Subject	SEM with categorical DV?

Hello:
 
I’m trying to test a mediational model in which all of the predictors (including the mediator) are continuous. The dependent variable, however, is dichotomous.  So far, I have run linear regressions to test predictors of the mediator and then I’ve used binary logistic regression for everything else. I’m concerned that testing the model in this way  (i.e., in halves, using 2 types of regression) introduces more problems than it’s worth. My understanding is that SEM requires a continuous DV, so I’m not sure what  type of analysis would permit me to analyze the model as a whole. I’d appreciate any suggestions (and thoughts on the validity/advisability of testing my model in the way I described).
Thanks so much
Joe
 
 
 
Joseph A. Bianco, Ph.D.
Department of Geriatric Medicine/Gerontology
Ohio University College of Osteopathic Medicine
353 Grosvenor
Athens, OH 45701
[phone]
[hidden email]