CFA using FIML with high missing rates in OpenMx?

1 reply [Last post]
cotto's picture
Offline
Joined: 09/04/2012

Hello, my name is christiane and I am a new user of Open MX.
In our study, we are looking for possibilites to analyse a combined data set with lots of missing data (we have combined data from 4 studies, just a few of our items were administered in each study). Can anyone tell me, if there is a covariance coverage limit implemeted in Open MX for CFA with Full-Information Maximum-likelihood (FIML)? Or how does Open MX deal with this?
Thank you and best wishes,
christiane

Ryne's picture
Offline
Joined: 07/31/2009
There is no lower (or upper)

There is no lower (or upper) limit for covariance coverage in OpenMx. The likelihood of each row is calculated using whatever data is available for that row. As long as some data is available for a given parameter, you shouldn't run into explicit problems. You may, however, have very wide standard errors or otherwise unstable runs when very little data is present for a given covariance.