Pandas dataframe aggregation changes type of boolean columns

Question

I am working with pandas data frames that have multiple columns, each of which can be numerical, categorical or boolean. In the mwe, the categorical column is a gene, and the values in 'bool_col' are either True or False for each gene. With pandas 1.2.4, aggregating on a categorical column using the median function works as I would expect:

import pandas as pd
df = pd.DataFrame.from_dict(
    {
        "A1": {"gene": "A", "bool_col": True, "S2": 10},
        "A2": {"gene": "A", "bool_col": True, "S2": 11},
        "B1": {"gene": "B", "bool_col": False, "S2": 40},
        "B2": {"gene": "B", "bool_col": False, "S2": 39},
        "C1": {"gene": "C", "bool_col": True, "S2": 9},
        "C2": {"gene": "C", "bool_col": True, "S2": 11},
    },
    orient="index"
)
df_genes = df.groupby("gene").agg("median")

As I expect, the aggregated data frame looks like this, the data type of the 'bool_col' column is still boolean and column 'S2' contains the median value of the respective category:

      bool_col    S2
gene                
A         True  10.5
B        False  39.5
C         True  10.0

However, after updating to pandas 1.3.4, the boolean columns are cast to the 'float64' data type during the aggregation:

      bool_col    S2
gene                
A          1.0  10.5
B          0.0  39.5
C          1.0  10.0

One could probably change the data type of the 'bool_col' column back to boolean, but is this behavior expected with pandas 1.3.4? Maybe I am missing something, but is there a way to keep the original boolean data type? Links and hints are highly appreciated!

Pandas dataframe aggregation changes type of boolean columns

0 Answers