Stata module to fit marginal logistic regression by inverse probability weighting
ipwlogit
fits marginal logistic regression of a binary dependent variable on
a treatment variable, possibly adjusting for control variables by inverse
probability weighting (IPW). The resulting estimate can be interpreted as a
marginal (log) odds ratio of a positive outcome. The treatment variable can be
categorical, continuous, or discrete.
Installation from GitHub:
. net from https://raw.githubusercontent.com/benjann/ipwlogit/main/
. net install ipwlogit, replace
ipwlogit
with option psmethod(gologit)
requires
gologit2
by Richard Williams. Type
. ssc install gologit2
to install gologit2
.
Main changes:
16jan2023 (version 1.0.8)
- support for vce(svy) added
- iweights no longer imply -novceadj-
- vce(bootstrap) and vce(jackknife) did not pick up level(); this is fixed
20oct2022 (version 1.0.7)
- truncate() is now applied to the overall distribution of stabilized weights
across all treatment levels, i.e. truncate() is no longer applied to each
treatment levels individually; non-stabilized weights are computed after
applying truncate()
- e(ipwstats) renamed to e(ipw)
- marginal treatment probabilities now returned in e(prop)
20oct2022 (version 1.0.6)
- option truncate() added
04sep2022 (version 1.0.5)
- tvar may now contain polynomials; parsing of varlist improved
- for categorical tvar, IPWs will now always be based on the observed levels,
not the levels specified in tvar
- options -rifgenerate()- and -ifscaling()- added
- e(sum_w) added to returns
01sep2022 (version 1.0.4)
- now using Sturges' rule to determine the number of bins used to categorize
a continuous treatment
17aug2022 (version 1.0.3)
- now requires Stata 14 or newer
- ttype "categorical" renamed to "factor"
- some adjustments to header display
16aug2022 (version 1.0.2)
- option nodots added
15aug2022 (version 1.0.1)
- fweights now supported
15aug2022 (version 1.0.0):
- ipwlogit released on GitHub