BUG: Remove change of global pickle method

Refactor PCA and Kernel methods to avoid instance methods
Remove buggy pickling method change in functional graphics

closes statsmodels#4772

statsmodels/graphics/functional.py

@@ -1,13 +1,13 @@
 """Module for functional boxplots."""
+from statsmodels.compat.python import range, zip
+from statsmodels.compat.scipy import factorial
 
 from statsmodels.multivariate.pca import PCA
 from statsmodels.nonparametric.kernel_density import KDEMultivariate
-from statsmodels.compat.python import range, zip
 from statsmodels.graphics.utils import _import_mpl
 from collections import OrderedDict
 from itertools import combinations
 import numpy as np
-from statsmodels.compat.scipy import factorial
 try:
  from scipy.optimize import differential_evolution, brute, fmin
  have_de_optim = True
@@ -16,31 +16,13 @@
  have_de_optim = False
 from multiprocessing import Pool
 import itertools
-try:
- import copyreg
-except ImportError:
- import copy_reg as copyreg
-import types
-
 from . import utils
 
 
 __all__ = ['hdrboxplot', 'fboxplot', 'rainbowplot', 'banddepth']
 
 
-def _pickle_method(m):
- """Handle pickling issues with class instance."""
- if m.im_self is None:
- return getattr, (m.im_class, m.im_func.func_name)
- else:
- return getattr, (m.im_self, m.im_func.func_name)
-
-
-copyreg.pickle(types.MethodType, _pickle_method)
-
-
 class HdrResults(object):
-
  """Wrap results and pretty print them."""
 
  def __init__(self, kwds):

statsmodels/multivariate/pca.py

@@ -229,14 +229,9 @@ def __init__(self, data, ncomp=None, standardize=True, demean=True,
  self._ncomp = min_dim
 
  self._method = method
- if self._method == 'eig':
- self._compute_eig = self._compute_using_eig
- elif self._method == 'svd':
- self._compute_eig = self._compute_using_svd
- elif self._method == 'nipals':
- self._compute_eig = self._compute_using_nipals
- else:
- raise ValueError('method is not known.')
+ # Workaround to avoid instance methods in __dict__
+ if self._method not in ('eig', 'svd', 'nipals'):
+ raise ValueError('method {0} is not known.'.format(method))
 
  self.rows = np.arange(self._nobs)
  self.cols = np.arange(self._nvar)
@@ -409,6 +404,19 @@ def _prepare_data(self):
  data = adj_data
  return data / np.sqrt(self.weights)
 
+ def _compute_eig(self):
+ """
+ Wrapper for actual eigenvalue method
+
+ This is a workaround to avoid instance methods in __dict__
+ """
+ if self._method == 'eig':
+ return self._compute_using_eig()
+ elif self._method == 'svd':
+ return self._compute_using_svd()
+ else: # self._method == 'nipals'
+ return self._compute_using_nipals()
+
  def _compute_using_svd(self):
  """SVD method to compute eigenvalues and eigenvecs"""
  x = self.transformed_data

statsmodels/nonparametric/_kernel_base.py

@@ -119,9 +119,6 @@ def _compute_bw(self, bw):
  -----
  The default values for bw is 'normal_reference'.
  """
-
- self.bw_func = dict(normal_reference=self._normal_reference,
- cv_ml=self._cv_ml, cv_ls=self._cv_ls)
  if bw is None:
  bw = 'normal_reference'
 
@@ -131,7 +128,13 @@ def _compute_bw(self, bw):
  else:
  # The user specified a bandwidth selection method
  self._bw_method = bw
- bwfunc = self.bw_func[bw]
+ # Workaround to avoid instance methods in __dict__
+ if bw == 'normal_reference':
+ bwfunc = self._normal_reference
+ elif bw == 'cv_ml':
+ bwfunc = self._cv_ml
+ else: # bw == 'cv_ls'
+ bwfunc = self._cv_ls
  res = bwfunc()
 
  return res

statsmodels/nonparametric/kernel_regression.py

@@ -95,7 +95,6 @@ def __init__(self, endog, exog, var_type, reg_type='ll', bw='cv_ls',
  self.exog = _adjust_shape(exog, self.k_vars)
  self.data = np.column_stack((self.endog, self.exog))
  self.nobs = np.shape(self.exog)[0]
- self.bw_func = dict(cv_ls=self.cv_loo, aic=self.aic_hurvich)
  self.est = dict(lc=self._est_loc_constant, ll=self._est_loc_linear)
  defaults = EstimatorSettings() if defaults is None else defaults
  self._set_defaults(defaults)
@@ -116,7 +115,11 @@ def _compute_reg_bw(self, bw):
  else:
  # The user specified a bandwidth selection method e.g. 'cv_ls'
  self._bw_method = bw
- res = self.bw_func[bw]
+ # Workaround to avoid instance methods in __dict__
+ if bw == 'cv_ls':
+ res = self.cv_loo
+ else: # bw == 'aic'
+ res = self.aic_hurvich
  X = np.std(self.exog, axis=0)
  h0 = 1.06 * X * \
  self.nobs ** (- 1. / (4 + np.size(self.exog, axis=1)))
@@ -494,7 +497,6 @@ def __init__(self, endog, exog, var_type, reg_type, bw='cv_ls',
  self.exog = _adjust_shape(exog, self.k_vars)
  self.data = np.column_stack((self.endog, self.exog))
  self.nobs = np.shape(self.exog)[0]
- self.bw_func = dict(cv_ls=self.cv_loo, aic=self.aic_hurvich)
  self.est = dict(lc=self._est_loc_constant, ll=self._est_loc_linear)
  defaults = EstimatorSettings() if defaults is None else defaults
  self._set_defaults(defaults)