Recently, I was trying to use Zelig to calculate ATEs from a
straightforward logit model. In other words, rather than calculating
first differences as a marginal effect at the means (i.e., for a
respondent whose scores on the covariates are equal to the sample means
of those variables), I wanted the quantity calculated as the average
difference in predicted probability across all respondents, using their
actual values of the covariates. Although the documentation says that I
should be able to get such a quantity through s.out$qi$ate.pr or
s.out$qi$ate.ev, I couldn't extract those quantities. At first I
thought that this was just an oversight peculiar to the logit model, but
it was also the case for a mixed-effects logit. The only elements that
were attached to my sim() object were:
[1] "x" "x1" "call"
"zelig.call" "par"
[6] "qi$ev" "qi$pr" "qi$rr"
"qi$fd"
Is there any chance that the ATE quantities were accidentally dropped
from a recent version, or am I doing something wrong?
Thanks,
Mike Hoffman
Show replies by date