Dear all,
a statistician friend of mine was describing certain well-known problems
with the -xtprobit- and -xtlogit- routines when panels are long.
Stata's -xt- manual discusses this as well, having to do with the
quadrature method used to implement the random-effects estimator. My
friend provided a "fix" using lowess techniques, and also mentioned a
different user-written ad-on with the file name "btscs" that he said was
difficult to find. Indeed, -findit- finds no trace. Does anyone have
access to the btscs program?
This question has come up before: