Panel-data models with random effects can be fit with Stata's me
commands for multilevel modeling. And the new metobit
command can fit panel-data tobit models to censored outcomes. For instance, if y is left-censored at 10, you could type:
to fit a model with random intercepts by id. In fact, you could fit this model with the existing xttobit
command.
What you cannot do with xttobit
is allow the slopes to vary by id. With metobit
, we include random slopes for x1 in addition to the random intercepts by typing:
You can see Multilevel tobit models for more information on metobit
and for an example with both random slopes and random intercepts.