AttributeError: 'pandas._libs.properties.CachedProperty' object has no attribute 'func'

[matteogabba]

When I run the following function

ar_select_order(seg, ic='aic', maxlag=maxlag)

I got this error message:

AttributeError: 'pandas._libs.properties.CachedProperty' object has no attribute 'func'

I never had this error before.

I run statsmodels version 0.13.5, pandas 1.4.4, scipy 1.9.3, on python 3.10.8 - Windows 11.

It seems to be related to this issue: https://github.com/pandas-dev/pandas/issues/46474

Example code to reproduce the issue:

import numpy as np
from statsmodels.tsa.ar_model import ar_select_order

x = np.array([ 239.83424321,  167.82769927,   89.8979825 ,   17.62647364,
        -74.61471784, -191.46196454, -317.49363405, -433.59813467,
       -489.33239431, -455.19897332, -353.48805058, -204.69413008,
        -32.88659459,  117.75119101,  256.05341303,  397.52653686,
        515.63154499,  582.87703054,  583.32171956,  487.0195182 ,
        314.68294874,  125.31024905,  -30.53799281, -116.31588371,
       -126.40961168,  -89.9293596 ,  -48.88184062,  -30.33868059,
        -50.98321497,  -97.06902425, -142.25422065, -160.52544434,
       -126.66740122,  -13.4300123 ,  160.02713276,  327.34313901,
        445.05741572,  442.63448766,  383.42265294,  696.5343972 ,
       1476.03014088, 2780.59407187,           np.nan,           np.nan,
                 np.nan,           np.nan,           np.nan,           np.nan,
                 np.nan,           np.nan])

idx_seg = [ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
        17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
        34, 35, 36, 37, 38, 39, 40, 41]

maxlag = 5

select = ar_select_order(seg, ic='aic', maxlag=maxlag)

Running import statsmodels.api as sm; sm.show_versions() raise the following error message:

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
Cell In [228], line 1
----> 1 import statsmodels.api as sm
      2 sm.show_versions()

File ~\anaconda3\envs\oop_test\lib\site-packages\statsmodels\api.py:124
    122 from .tools.tools import add_constant, categorical
    123 from .tools.web import webdoc
--> 124 from .tsa import api as tsa
    125 from ._version import get_versions
    127 __version__ = get_versions()["version"]

File ~\anaconda3\envs\oop_test\lib\site-packages\statsmodels\tsa\api.py:74
     72 from .base import datetools
     73 from .exponential_smoothing.ets import ETSModel
---> 74 from .filters import api as filters, bk_filter, cf_filter, hp_filter
     75 from .forecasting.stl import STLForecast
     76 from .holtwinters import ExponentialSmoothing, Holt, SimpleExpSmoothing

File ~\anaconda3\envs\oop_test\lib\site-packages\statsmodels\tsa\filters\api.py:6
      4 from .hp_filter import hpfilter
      5 from .cf_filter import cffilter
----> 6 from .filtertools import miso_lfilter, convolution_filter, recursive_filter

File ~\anaconda3\envs\oop_test\lib\site-packages\statsmodels\tsa\filters\filtertools.py:18
     16 import scipy.fftpack as fft
     17 from scipy import signal
---> 18 from scipy.signal.signaltools import _centered as trim_centered
     20 from statsmodels.tools.validation import array_like, PandasWrapper
     22 def _pad_nans(x, head=None, tail=None):

ImportError: cannot import name '_centered' from 'scipy.signal.signaltools' (C:\Users\voodo\anaconda3\envs\oop_test\lib\site-packages\scipy\signal\signaltools.py)

[bashtage]

Your scipy is too new. You need to downgrade it to something earlier.

ImportError: cannot import name '_centered' from 'scipy.signal.signaltools' (C:\Users\voodo\anaconda3\envs\oop_test\lib\site-packages\scipy\signal\signaltools.py)

[bashtage]

You might also need to downgrade your pandas.

[matteogabba]

Can you please suggest which versions should I use?

[bashtage]

1.7.x seems to be the latest that has that attribute.

[bashtage]

You don't have the statsmodels you think you have. This is not an issue in 0.13.5. You should upgrade to 0.13.5 which works with both of these (pandas and scipy).

[matteogabba]

OK. If I run conda list it indicates the latest version. What is the reason of it? I will anyway try to reinstall a fresh virtenv. I will keep you updated. Thanks!

[bashtage]

Hard to say, but there is a try, except ImportError in the 0.13.x branch, and we test against these versions of pandas and scipy.