That leads to me to ask a question: which would
we nominate as (say) the top ten tricks which
are the deepest and most Stataish features
in what we use? What is _both_ simple _and_ deep?
What leads to great results with at most a few
lines of code?
I very much like -predict-. I especially like that it can be run on
something other than the estimation sample. Indeed, I sometimes
temporarily wipe out the "real" data, type in some hypothetical values, run
-predict-, and then restore the original data. This can be quite useful
for making things like logistic regression more tangible, where it is hard
to see what impact variables actually have. I also like -adjust- for
similar reasons.
