Sometimes the is_continuous is not precise enough to identify continuous and categorical variables, leading to unwieldy plots. Also sometimes it's useful to only plot the balance on a few specific covariates of interest after matching.
Changes
In order to enable that, I have added a new argument columns: List[str] to the Matcher.compare_continuous and Matcher.compare_categorical methods. When columns is passed in, we do not verify if the column is continuous or categorical but we do remove columns in the self.exclude set.
There is no actual change within the plotting for loops (as indicated below). The changes are simply due to the removal of indentation.
# from
for col in self.matched_data.columns:
if uf.is_continuous(col, self.X) and col not in self.exclude:
do_stuff()
# to
for col in columns_to_plot:
do_stuff()
Verification
I ran the example notebook with and without columns and they work as intended. See nbviewer.
Context
Sometimes the
is_continuousis not precise enough to identify continuous and categorical variables, leading to unwieldy plots. Also sometimes it's useful to only plot the balance on a few specific covariates of interest after matching.Changes
In order to enable that, I have added a new argument
columns: List[str]to theMatcher.compare_continuousandMatcher.compare_categoricalmethods. Whencolumnsis passed in, we do not verify if the column is continuous or categorical but we do remove columns in theself.excludeset.There is no actual change within the plotting for loops (as indicated below). The changes are simply due to the removal of indentation.
Verification
I ran the example notebook with and without columns and they work as intended. See nbviewer.