

Corr btw binary exogenous & latent co... 

Message/Author 

Heili Pals posted on Wednesday, August 01, 2007  12:54 pm



I have a path model, where a binary exogenous variable is correlated with a continuous latent variable. Because the binary variable is exogenous, MPLUS does not allow me to define it to be categorical (using the CATEGORICAL ARE command). The output however, warns me that I have not defined my binary variable as binary. Is my model estimating the correlation correctly? Can I estimate a correlation between a binary exogenous variable and a continuous latent variable? The categorical exogenous variable under question in the model below is dt1viol. USEV ARE t1dep t1anx t1dero t3dep t3anx t3dero dt1viol dt3viol devid; CATEGORICAL ARE dt3viol t1anx t1dep t1dero t3anx t3dep t3dero devid; ANALYSIS: ESTIMATOR = wlsmv; PARAMETERIZATION = theta; MODEL: t1nsf BY t1anx t1dep t1dero; t3nsf BY t3anx t3dep t3dero; t1nsf WITH devid dt1viol; t1dero WITH dt1viol; devid ON dt1viol; t3nsf ON t1nsf devid ; dt3viol ON dt1viol devid t1nsf; dt3viol WITH t3nsf t3dero; t3anx WITH t1anx; 


With weighted least squares estimation, you do not want to correlate an exogenous observed categorical variable and an exogenous latent variable because to do so you must treat the exogenous observed variable as a dependent variable in the model. You should instead regress the factor on the categorical observed variable. 

Kihan Kim posted on Monday, March 14, 2011  1:25 am



Can I run SEM in Mplus when an exogenous variable is binary, and the endogenous variable is continuous? 


Yes. 

Back to top 

