I'm unclear about how Stata's GSEM handles missing values. I know that SEM uses full-information maximum likelihood (FIML), but GSEM does not have that option. However, the Stata documentation claims that GSEM is "sometimes able" to use more data than would otherwise be available through listwise deletion. Could anyone elaborate on what this "sometimes" actually means?
In the meantime, I have explored using multiple imputation in combination with GSEM (using the command "mi estimate, cmdok: gsem...") but this is pretty experimental and I would prefer not to go down this route.
In the meantime, I have explored using multiple imputation in combination with GSEM (using the command "mi estimate, cmdok: gsem...") but this is pretty experimental and I would prefer not to go down this route.
