missing = “fiml” not available in the categorical setting

[Jack Bailey]

Hi,

I use a lot of categorical data and recent versions of lavaan seem to have been able to handle missingness using fiml. I've just updated and get the following error when running models that used to work: 

"lavaan ERROR: missing = “fiml” not available in the categorical setting"

Is there a reason why it no longer works?

[Terrence Jorgensen]

Is there a reason why it no longer works?

It never did.  FIML is ML estimation, which assumes multivariate normality.  Perhaps the error message is the result of a new check for this inconsistency between the estimator and handling of missing data.

In the past, lavaan has had an experimental (i.e., not supported) feature for MLE of categorical data, which appears to have been turned off for now.  So the only estimator available for categorical data is (robust) DWLS.

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

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

http://www.uva.nl/profile/t.d.jorgensen

[Jack Bailey]

I reran the models using pairwise deletion and got the same results. Perhaps it was letting me set missing = "fiml", but still using pairwise deletion nonetheless.