Unexpected TypeError in remove_duplicate_columns

[NicolasGensollen]

• What is the current behavior?

When summing FixedEffect instances built from a Pandas DataFrame with converted dtypes, the function remove_duplicate_columns in brainstat/stats/terms.py raises the following error:

File ~/GitRepos/BrainStat/brainstat/stats/terms.py:243, in remove_duplicate_columns(df, tol)
    241 df *= 1 / tol
    242 # keep = df.round().T.drop_duplicates(keep="last").T.columns  # Slow!!
--> 243 idx = np.unique(df.round().values, axis=1, return_index=True)[-1]
    244 keep = df.columns[sorted(idx)]
    245 return keep

TypeError: The axis argument to unique is not supported for dtype object

• Please provide the steps to reproduce and if possible a minimal demo of the problem.

Here is a MWE:

import pandas as pd
from brainstat.stats.terms import FixedEffect

df = pd.DataFrame({"age": [20.2, 33.1, 25.0],
                   "sex": ["Female", "Male", "Male"],
                  }).convert_dtypes()
df = df.apply(lambda x: x.astype('category') if x.name == "sex" else x)
FixedEffect(df.age) + FixedEffect(df.sex)

• What is the expected behavior?

No crash.

• Please tell us about your computing environment:

Python 3.9.11
Brainstat 0.3.6

• Other information