If StataCorp are listening, I have no problem with adding "the
capability of saving multiple sets of windowing preference sets to the
Stata Wish list" in the abstract. But the workday is not infinite.
Hence, my preference would be for StataCorp to devote like 95% of the time
to extending the models estimable and extending the things one can do
after estimation. For example, multilevel models are not feasibly
estimated in Stata, despite the presence of gllamm. (While gllamm is
nice, it does not work well with datasets greater than, say, 10,000, and
it is not supported, while a module for multilevel modelling that is fully
part of stata would be supported). The ability to estimate such models
will make it possible for us to learn new things about the world. I hope
StataCorp prioritizes those things that will help us learn new facts about
the world above other possible priorities (e.g., tweaking the interface,
surely a complex programming task with comparatively little benefit
compared to an at least equally complex programming task of introducing
new models into stata).
--------------------------------------------------------
