BUG: .describe() doesn't work for EAs

[andrewgsavage]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd, pint_pandas
s = pd.Series([1, 2, 3], dtype='pint[kg]')
s.describe()

DimensionalityError                       Traceback (most recent call last)
...

Issue Description

hgrecco/pint-pandas#279

Series.describe sets the dtype for the results to Float64Dtype when the input is an EA. pint's Quantity
cannot be casted to Float64Dtype.

pandas/pandas/core/methods/describe.py

Line 255 in 35b0d1d

dtype = Float64Dtype()

Expected Behavior

.describe should return a Series of objectdtype, or the dtype of the EA

Installed Versions

Replace this line with the output of pd.show_versions()

[kernelism]

take

[andrewgsavage]

I suggest waiting for a maintainer response before working on this. The logic for the dtype is not trivial. For example, taking the standard deviation a temperature Series gives a delta temperature unit, while min, max etc have a temperature unit. These cannot be stored in the same dtype.

…

On Thu, Jun 26, 2025 at 12:42 PM Arjhun S ***@***.***> wrote: *kernelism* left a comment (pandas-dev/pandas#61707) <#61707 (comment)> take — Reply to this email directly, view it on GitHub <#61707 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADEMLEF6FGIVG25HF6SJCBT3FPMBFAVCNFSM6AAAAACAEQEDQKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAMBYGE4TAMJWGQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[kernelism]

@andrewgsavage I’ve been exploring the codebase (still new here), and my initial thought is to check if the list of calculated statistics contains multiple types. If so, setting dtype=None would cause the result to be a Series with object dtype, which should resolve the issue. Does that sound right?

[andrewgsavage]

I think that is a good solution. I wonder how it would deal with an Series with int dtype. Should that give objects or float dtype?, since mean would give a float while mean would give int

[kernelism]

I think that is a good solution. I wonder how it would deal with an Series with int dtype. Should that give objects or float dtype?, since mean would give a float while mean would give int

In that case, it would never go into the if block checking if the series dtype is EA. It would automatically get a float type based on existing logic. Our fix would only pertain to EAs.

pandas/pandas/core/methods/describe.py

Lines 256 to 258 in 35b0d1d

	elif series.dtype.kind in "iufb":
	# i.e. numeric but exclude complex dtype
	dtype = np.dtype("float")

Should I open a PR?