When working with categorical variables, not using observed=True creates groups with no records. This is unintuitive and causing tons of headaches. We set observed=True manually pretty much everywhere, so it'd be nice if this was the default.
Pandas also throws a warning saying that observed=True will be default in the future.
When working with categorical variables, not using
observed=Truecreates groups with no records. This is unintuitive and causing tons of headaches. We setobserved=Truemanually pretty much everywhere, so it'd be nice if this was the default.Pandas also throws a warning saying that
observed=Truewill be default in the future.