Special Features: Estimation Methods and Other Stuff

Estimation methods

Typicall, Mplus will decide on its own which estimation method to use. It will do so depending on the type of data (e. g., whether variables are metric only or not), or the presence of missing values, or other features. Yet, sometimes you may wish to request explicitly a specific method of estimation. This is achieved via the ANALYSIS command, such as in:

ANALYSIS:
ESTIMATOR IS MLR;

which may of course supplemented by other subcommands.

Some of the methods available are:

ML	Standard maximum likelihood estimator
MLM	Maximum likelihood with robust standard errors (so-called Satorra-Bentler estimator)
MLR	Maximum likelihood with robust standard errors that also take clustering of cases into account

More about this is to follow.

Other features

Bootstrapping

ANALYSIS:
BOOTSTRAP = 1000;

will compute confidence intervals based on boostrapping (with the number of bootstraps indicated after the equals sign). Don't worry – this is very fast, at least with simple models (I have not yet tried complicated ones). Note that the confidence intervals based on bootstrapping have to be requested in the OUTPUT section.