i've been running some first differences in clarify
that are based on multinomial logit models where the
dependent variable has a base category 0 and
comparison categories -1 and 1. i've noticed that
the first difference for the highest category (in this
case, 1) always has a *tiny* standard error,
regardless of the error in the corresponding
coefficient. i can reverse the order of the variable
(1 becomes -1 and vice versa) and the highest category
still gets the smallest s.e. i can obviously code my
dep var to give the highest value to the comparison
category so the results i report are more in line with
the coefficients, but the phenomenon has concerned me.
does anyone know why this happens? does it have to
do with the fact that the results of any given
category are always perfectly predicted by the results
of the other two? or is it something about my data?
confused,
eric mcghee
__________________________________________________
Do You Yahoo!?
Yahoo! Finance - Get real-time stock quotes
http://finance.yahoo.com
clarify mailing list served by Harvard-MIT Data Center
List Address: clarify(a)latte.harvard.edu
Subscribe/Unsubscribe:
http://lists.hmdc.harvard.edu/listis.cgi?info=clarify