i am trying to run a logit regression with a fixed effect.
i tried doing it in two different ways:
the first time i generated a dummy for the fixed variable (99 dummies) and ran a regular logit regression
. quiely tab code, gen(code)
. logit sharap oprmin anmin elect los age sex reop24 reop13 reop37 reop7m code1-code98
the second time i used xtlogit
.xtlogit sharap oprmin anmin elect los age sex reop24 reop13 reop37 reop7m, fe
my first question is are these two methods equivalent. should i get similar results for the two.
if the answer is yes, why would i get an error message (Hessian is not negative semidefinite) in the second mehod but not the first.
if the answer is no, i would appreciate if someone could explain why there is a difference
can someone explain what the error "Hessian is not negative semidefinite" meens and what i can do to fix it?