pandas NaN defaults: all-NaN groups sum to 0, NaN keys vanish from totals

Same results on pandas 2.2.3 and 3.1.0. With g=['a','a',NaN,'b'] and v=[1,2,10,NaN]: (1) Series([nan, nan]).sum() is 0.0 and prod() is 1.0, while mean/max are NaN, so groupby('g').v.sum() reports b=0.0 for a group with no data. Pass min_count=1 to get NaN. (2) groupby drops NaN keys by default, so the group sums add up to 3.0 while v.sum() is 13.0. Use groupby(..., dropna=False); pivot_table(dropna=False) also keeps the NaN row. (3) value_counts(normalize=True) divides by the non-NaN count (a=0.667, b=0.333 for 4 rows); with dropna=False it is a=0.5, NaN=0.25, b=0.25. A cheap check after any groupby: assert the group totals equal the column total.