On Tue, 7 May 2002, Paul von Hippel wrote:
relogit has two weighting options:
(1) a general-purpose "weight" option like that used by many stata commands;
(2) a "wc" option to compensate for oversampling cases where Y=1.
General question: What happens if I use (1) and (2) together? What happens
if I use (1) alone? What happens if the values in (1) conflict with (2)?
the "wc" option weights the 1's in the data by t/m where t is the
population fraction of 1's and y the sample fraction, and weights the 0's
in the data by (1-t)/(1-m). so it's equivalent to the "weight" option if
your weighting variable for that option contains these same values for the
corresponding observations.
so for the purpose of correcting oversampling of y=1 cases, you can use
either of these but not both. "wc" doesn't correct other types of
non-random sampling (such as those based on values of the x's).
Specific question: In my data, I have a weight variable that compensates
both for oversampling cases where Y=1 and for oversampling cases with
certain values of X.
I'm inclined to think that I should use this variable with the "weight"
option and ignore the "wc" option. Will this provide correct results?
i think so.
Langche Zeng
Many thanks,
Paul von Hippel
-
relogit mailing list served by Harvard-MIT Data Center
List Address: relogit(a)latte.harvard.edu
Subscribe/Unsubscribe:
http://lists.hmdc.harvard.edu/?info=relogit
-
relogit mailing list served by Harvard-MIT Data Center
List Address: relogit(a)latte.harvard.edu
Subscribe/Unsubscribe:
http://lists.hmdc.harvard.edu/?info=relogit