MAINT: Relax test tolerance for future compat

Relax very tight test coverage

statsmodels/base/model.py

@@ -1,6 +1,7 @@
 from statsmodels.compat.python import lzip
 
 from functools import reduce
+import warnings
 
 import numpy as np
 from scipy import stats
@@ -543,9 +544,8 @@ def hess(params, *args):
                 Hinv = eigvecs.dot(np.diag(1.0 / eigvals)).dot(eigvecs.T)
                 Hinv = np.asfortranarray((Hinv + Hinv.T) / 2.0)
             else:
-                from warnings import warn
-                warn('Inverting hessian failed, no bse or cov_params '
-                     'available', HessianInversionWarning)
+                warnings.warn('Inverting hessian failed, no bse or cov_params '
+                              'available', HessianInversionWarning)
                 Hinv = None
 
         if 'cov_type' in kwargs:
@@ -562,10 +562,10 @@ def hess(params, *args):
         mlefit.mle_retvals = retvals
         if isinstance(retvals, dict):
             if warn_convergence and not retvals['converged']:
-                from warnings import warn
                 from statsmodels.tools.sm_exceptions import ConvergenceWarning
-                warn("Maximum Likelihood optimization failed to converge. "
-                     "Check mle_retvals", ConvergenceWarning)
+                warnings.warn("Maximum Likelihood optimization failed to "
+                              "converge. Check mle_retvals",
+                              ConvergenceWarning)
 
         mlefit.mle_settings = optim_settings
         return mlefit
@@ -967,7 +967,6 @@ def fit(self, start_params=None, method='nm', maxiter=500, full_output=1,
                                               for i in range(-k_miss)])
             else:
                 # I do not want to raise after we have already fit()
-                import warnings
                 warnings.warn('more exog_names than parameters', ValueWarning)
 
         return genericmlefit
@@ -1084,7 +1083,6 @@ def predict(self, exog=None, transform=True, *args, **kwargs):
                        '{0}'.format(str(str(exc))))
                 raise exc.__class__(msg)
             if orig_exog_len > len(exog) and not is_dict:
-                import warnings
                 if exog_index is None:
                     warnings.warn('nan values have been dropped', ValueWarning)
                 else:
@@ -1353,24 +1351,30 @@ def bse(self):
             bse_ = np.empty(len(self.params))
             bse_[:] = np.nan
         else:
-            bse_ = np.sqrt(np.diag(self.cov_params()))
+            with warnings.catch_warnings():
+                warnings.simplefilter("ignore", RuntimeWarning)
+                bse_ = np.sqrt(np.diag(self.cov_params()))
         return bse_
 
     @cached_value
     def tvalues(self):
         """
         Return the t-statistic for a given parameter estimate.
         """
-        return self.params / self.bse
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore", RuntimeWarning)
+            return self.params / self.bse
 
     @cached_value
     def pvalues(self):
         """The two-tailed p values for the t-stats of the params."""
-        if self.use_t:
-            df_resid = getattr(self, 'df_resid_inference', self.df_resid)
-            return stats.t.sf(np.abs(self.tvalues), df_resid) * 2
-        else:
-            return stats.norm.sf(np.abs(self.tvalues)) * 2
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore", RuntimeWarning)
+            if self.use_t:
+                df_resid = getattr(self, 'df_resid_inference', self.df_resid)
+                return stats.t.sf(np.abs(self.tvalues), df_resid) * 2
+            else:
+                return stats.norm.sf(np.abs(self.tvalues)) * 2
 
     def cov_params(self, r_matrix=None, column=None, scale=None, cov_p=None,
                    other=None):
@@ -1563,7 +1567,6 @@ def t_test(self, r_matrix, cov_p=None, scale=None, use_t=None):
         ==============================================================================
         """
         if scale is not None:
-            import warnings
             warnings.warn('scale is has no effect and is deprecated. It will'
                           'be removed in the next version.',
                           DeprecationWarning)
@@ -1715,7 +1718,6 @@ def f_test(self, r_matrix, cov_p=None, scale=1.0, invcov=None):
         <F test: F=array([[ 144.17976065]]), p=6.322026217355609e-08, df_denom=9, df_num=3>
         """
         if scale != 1.0:
-            import warnings
             warnings.warn('scale is has no effect and is deprecated. It will'
                           'be removed in the next version.',
                           DeprecationWarning)
@@ -1786,7 +1788,6 @@ def wald_test(self, r_matrix, cov_p=None, scale=1.0, invcov=None,
         where the rank of the covariance of the noise is not full.
         """
         if scale != 1.0:
-            import warnings
             warnings.warn('scale is has no effect and is deprecated. It will'
                           'be removed in the next version.',
                           DeprecationWarning)
@@ -1828,7 +1829,6 @@ def wald_test(self, r_matrix, cov_p=None, scale=1.0, invcov=None,
             invcov = np.linalg.pinv(cov_p)
             J_ = np.linalg.matrix_rank(cov_p)
             if J_ < J:
-                import warnings
                 warnings.warn('covariance of constraints does not have full '
                               'rank. The number of constraints is %d, but '
                               'rank is %d' % (J, J_), ValueWarning)

statsmodels/base/tests/test_generic_methods.py

@@ -109,19 +109,14 @@ def test_zero_constrained(self):
         assert_equal(res1.bse[drop_index], 0)
         # OSX has many slight failures on this test
         tol = 1e-8 if PLATFORM_OSX else 1e-10
-        with pytest.warns(RuntimeWarning, match="invalid value encountered"):
-            # division by zero in bse
-            tvals1 = res1.tvalues[keep_index_p]
+        tvals1 = res1.tvalues[keep_index_p]
         assert_allclose(tvals1, res2.tvalues, rtol=tol, atol=tol)
 
         # See gh5993
         if PLATFORM_LINUX32 or SCIPY_GT_14:
             pvals1 = res1.pvalues[keep_index_p]
         else:
-            with pytest.warns(RuntimeWarning,
-                              match="invalid value encountered"):
-                # passing NaN into scipy.stats functions
-                pvals1 = res1.pvalues[keep_index_p]
+            pvals1 = res1.pvalues[keep_index_p]
         assert_allclose(pvals1, res2.pvalues, rtol=tol, atol=tol)
 
         if hasattr(res1, 'resid'):
@@ -259,19 +254,14 @@ def test_zero_collinear(self):
             assert_allclose(res1.params[drop_index], 0, rtol=1e-10)
             assert_allclose(res1.bse[keep_index_p], res2.bse, rtol=1e-8)
             assert_allclose(res1.bse[drop_index], 0, rtol=1e-10)
-            with pytest.warns(RuntimeWarning, match="invalid value"):
-                # zero in bse, so division by zero warning
-                tvals1 = res1.tvalues[keep_index_p]
+            tvals1 = res1.tvalues[keep_index_p]
             assert_allclose(tvals1, res2.tvalues, rtol=5e-8)
 
             # See gh5993
             if PLATFORM_LINUX32 or SCIPY_GT_14:
                 pvals1 = res1.pvalues[keep_index_p]
             else:
-                with pytest.warns(RuntimeWarning,
-                                  match="invalid value encountered"):
-                    # passing NaN into scipy.stats functions
-                    pvals1 = res1.pvalues[keep_index_p]
+                pvals1 = res1.pvalues[keep_index_p]
             assert_allclose(pvals1, res2.pvalues, rtol=1e-6, atol=1e-30)
 
             if hasattr(res1, 'resid'):

statsmodels/nonparametric/linbin.pyx

@@ -20,7 +20,7 @@ def fast_linbin(np.ndarray[DOUBLE] X, double a, double b, int M, int trunc=1):
         Py_ssize_t i, li_i
         int nobs = X.shape[0]
         double delta = (b - a)/(M - 1)
-        np.ndarray[DOUBLE] gcnts = np.zeros(M, np.float)
+        np.ndarray[DOUBLE] gcnts = np.zeros(M, float)
         np.ndarray[DOUBLE] lxi = (X - a)/delta
         np.ndarray[INT] li = lxi.astype(int)
         np.ndarray[DOUBLE] rem = lxi - li

statsmodels/regression/dimred.py

@@ -1,8 +1,11 @@
 import warnings
+
 import numpy as np
 import pandas as pd
+
 from statsmodels.base import model
 import statsmodels.base.wrapper as wrap
+from statsmodels.tools.sm_exceptions import ConvergenceWarning
 
 
 class _DimReductionRegression(model.Model):
@@ -688,7 +691,7 @@ def fit(self, start_params=None, maxiter=200, gtol=1e-4):
             g = self.score(params.ravel())
             gn = np.sqrt(np.sum(g * g))
             msg = "CovReduce optimization did not converge, |g|=%f" % gn
-            warnings.warn(msg)
+            warnings.warn(msg, ConvergenceWarning)
 
         results = DimReductionResults(self, params, eigs=None)
         results.llf = llf

statsmodels/stats/dist_dependence_measures.py

@@ -14,13 +14,14 @@
    Annals of Statistics, Vol. 35 No. 6, pp. 2769-2794.
 
 """
-import numpy as np
-import warnings
 from collections import namedtuple
+import warnings
 
+import numpy as np
 from scipy.spatial.distance import pdist, squareform
 from scipy.stats import norm
 
+from statsmodels.tools.sm_exceptions import HypothesisTestWarning
 
 DistDependStat = namedtuple(
     "DistDependStat",
@@ -131,10 +132,10 @@ def distance_covariance_test(x, y, B=None, method="auto"):
     # to the asymptotic approximation.
     if chosen_method == "emp" and pval in [0, 1]:
         msg = (
-            "p-value was {} when using the empirical method. ".format(pval)
-            + "The asymptotic approximation will be used instead"
+            f"p-value was {pval} when using the empirical method. "
+            "The asymptotic approximation will be used instead"
         )
-        warnings.warn(msg)
+        warnings.warn(msg, HypothesisTestWarning)
         _, pval = _asymptotic_pvalue(stats)
 
     return test_statistic, pval, chosen_method