conda环境下statsmodels子模块导入失败及安装报错问题咨询

Hey there, let's break down your issues and fix them step by step:

What's Going On Here? #

You've got two separate problems: a DLL load failure when importing statsmodels, and a PackagesNotFoundError when trying to install submodules that don't exist as standalone packages.

Your Import Code #

import numpy as np
import statsmodels.api as statm
import statsmodels.formula.api as statfa
import pandas as pd
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

The Error Traceback #

Traceback (most recent call last):
File "C:/Users/aboi/PycharmProjects/deeplearning/deep_class/05_3D_Graph.py", line 2, in <module>
import statsmodels.api as statm
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\statsmodels\api.py", line 7, in <module>
from . import regression
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\statsmodels\regression\__init__.py", line 1, in <module>
from .linear_model import yule_walker
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\statsmodels\regression\linear_model.py", line 40, in <module>
from scipy import stats
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\stats\__init__.py", line 379, in <module>
from .stats import *
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\stats\stats.py", line 182, in <module>
from . import distributions
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\stats\distributions.py", line 10, in <module>
from ._distn_infrastructure import (entropy, rv_discrete, rv_continuous,
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\stats\_distn_infrastructure.py", line 24, in <module>
from scipy import optimize
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\optimize\__init__.py", line 391, in <module>
from ._minimize import *
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\optimize\_minimize.py", line 30, in <module>
from ._trustregion_constr import _minimize_trustregion_constr
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\optimize\_trustregion_constr\__init__.py", line 4, in <module>
from .minimize_trustregion_constr import _minimize_trustregion_constr
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\optimize\_trustregion_constr\minimize_trustregion_constr.py", line 4, in <module>
from scipy.sparse.linalg import LinearOperator
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\sparse\linalg\__init__.py", line 116, in <module>
from .eigen import *
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\sparse\linalg\eigen\__init__.py", line 11, in <module>
from .arpack import *
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\sparse\linalg\eigen\arpack\__init__.py", line 22, in <module>
from .arpack import *
File "C:\Users\aboi\.conda\envs\hw\lib\site-packages\scipy\sparse\linalg\eigen\arpack\arpack.py", line 45, in <module>
from . import _arpack
ImportError: DLL load failed: dll load failed the specified procedure could not be found.

The Extra Package Install Error #

When you ran conda install statsmodels.api and conda install statsmodels.formula.api, you got a PackagesNotFoundError because those aren't real conda packages—they're just submodules of the main statsmodels library.

Why This Happens #

1. DLL Load Failure: The root issue is with scipy's ARPACK module. This usually happens because:

   • Your scipy version is incompatible with other dependencies (like numpy or MKL) in your conda environment
   • The pre-built scipy package you installed is missing required system DLLs
   • There are conflicting versions of scipy or its dependencies floating around in your environment

2. Invalid Package Names: statsmodels.api and statsmodels.formula.api aren't standalone packages—you don't install them separately. They come included with the main statsmodels package once it's properly installed.

Fixes to Try #

First: Fix the Scipy/Statsmodels Compatibility #

The most reliable fix is to reinstall scipy and statsmodels using a channel that ensures compatibility:

1. Activate your conda environment first:

   conda activate hw
   

2. Uninstall the problematic versions completely:

   conda remove --force scipy statsmodels
   

3. Install using the conda-forge channel (their packages are better tested for cross-dependency compatibility):

   conda install -c conda-forge scipy statsmodels numpy pandas matplotlib
   

Second: Verify the Fix #

Fire up a Python shell in your activated environment and run your import code again:

import numpy as np
import statsmodels.api as statm
import statsmodels.formula.api as statfa
import pandas as pd
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
print("All modules imported successfully!")

If That Still Doesn't Work #

• Install VC++ Redistributable: Sometimes DLL errors are caused by missing system libraries. Grab the 64-bit Microsoft Visual C++ Redistributable for Visual Studio 2019 and install it.
• Create a Fresh Environment: Old environments can get messy with conflicting dependencies. Start fresh:

  conda create -n new_hw python=3.9 numpy pandas matplotlib scipy statsmodels -c conda-forge
  conda activate new_hw
  

  (Python 3.9 or 3.10 are stable choices with great package compatibility.)

A Quick Note on Submodules #

Remember: You never need to install statsmodels.api or statsmodels.formula.api directly. As long as the main statsmodels package is installed correctly, those submodules are available to import right away.

内容的提问来源于stack exchange，提问作者RainOver