Modelling Residuals for Ordered Variables when conditional.x=TRUE

11 views

Skip to first unread message

cca...@broadinstitute.org

unread,

Feb 25, 2019, 5:47:46 PM2/25/19

to lavaan

Hi,

I am using lavaan for a CFA that has a mix of binary, ordinal, and continuous endogenous variables, as well as multiple continuous (age, etc) and one binary (sex) exogenous variable. I would like to use the option conditional.x=TRUE in order to partial out the effects of the exogenous covariates prior to modelling the factors, but I want to verify that the residuals will still be modelled correctly for the ordered variables if they are conditioned first on the exogenous variables. Could you please let me know how they are modelled in this case?

Thanks,

Caitlin

Terrence Jorgensen

unread,

Feb 28, 2019, 10:14:06 PM2/28/19

to lavaan

I want to verify that the residuals will still be modelled correctly for the ordered variables if they are conditioned first on the exogenous variables.

what do you mean by "correctly"?

Could you please let me know how they are modelled in this case?

The estimated residual polychoric correlation matrix is used in place of the estimated zero-order polychoric correlation matrix, which itself is already used in place of the "observed" covariance matrix (for continuous data) typically denoted as "S" in the discrepancy function depicted in SEM textbooks.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam