Merge 3f58ac3 into 5c7d4be

statsmodels/base/model.py

@@ -478,7 +478,6 @@ def __init__(self, endog, exog=None, loglike=None, score=None,
             self.score = score
         if not hessian is None:
             self.hessian = hessian
-        self.confint_dist = stats.norm
 
         self.__dict__.update(kwds)
 
@@ -854,11 +853,16 @@ class LikelihoodModelResults(Results):
             allvecs : list
                 Results at each iteration.
         """
+
+    # by default we use normal distribution
+    # can be overwritten by instances or subclasses
+    use_t = False
+
     def __init__(self, model, params, normalized_cov_params=None, scale=1.):
         super(LikelihoodModelResults, self).__init__(model, params)
         self.normalized_cov_params = normalized_cov_params
         self.scale = scale
-        self.use_t = False  # by default we use normal distribution
+
 
     def normalized_cov_params(self):
         raise NotImplementedError
@@ -880,7 +884,11 @@ def tvalues(self):
 
     @cache_readonly
     def pvalues(self):
-        return stats.norm.sf(np.abs(self.tvalues)) * 2
+        if self.use_t:
+            return stats.t.sf(np.abs(self.tvalues), self.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):
@@ -1379,8 +1387,13 @@ def conf_int(self, alpha=.05, cols=None, method='default'):
         method.
         """
         bse = self.bse
-        dist = stats.norm
-        q = dist.ppf(1 - alpha / 2)
+
+        if self.use_t:
+            dist = stats.t
+            q = dist.ppf(1 - alpha / 2, self.df_resid)
+        else:
+            dist = stats.norm
+            q = dist.ppf(1 - alpha / 2)
 
         if cols is None:
             lower = self.params - q * bse

statsmodels/base/tests/test_generic_methods.py

@@ -85,6 +85,12 @@ def test_ftest_pvalues(self):
                                                    for k in range(k_vars)]
         assert_allclose(pvals, res.pvalues, rtol=5e-10, atol=1e-25)
 
+        # label for pvalues in summary
+        string_use_t = 'P>|z|' if use_t is False else 'P>|t|'
+        summ = str(res.summary())
+        assert_(string_use_t in summ)
+        summ = str(res.summary2())
+        assert_(string_use_t in summ)
 
 
     # TODO The following is not (yet) guaranteed across models

statsmodels/discrete/discrete_model.py

@@ -2394,7 +2394,7 @@ def summary(self, yname=None, xname=None, title=None, alpha=.05,
                           yname=yname, xname=xname, title=title)
         # for parameters, etc
         smry.add_table_params(self, yname=yname_list, xname=xname, alpha=alpha,
-                             use_t=False)
+                             use_t=self.use_t)
 
         if hasattr(self, 'constraints'):
             smry.add_extra_txt(['Model has been estimated subject to linear '

statsmodels/genmod/generalized_linear_model.py

@@ -1046,7 +1046,7 @@ def summary(self, yname=None, xname=None, title=None, alpha=.05):
         smry.add_table_2cols(self, gleft=top_left, gright=top_right,  # [],
                              yname=yname, xname=xname, title=title)
         smry.add_table_params(self, yname=yname, xname=xname, alpha=alpha,
-                              use_t=True)
+                              use_t=self.use_t)
 
         if hasattr(self, 'constraints'):
             smry.add_extra_txt(['Model has been estimated subject to linear '

statsmodels/iolib/summary2.py

@@ -133,7 +133,7 @@ def add_base(self, results, alpha=0.05, float_format="%.4f", title=None,
             Name of the dependent variable (optional)
         '''
 
-        param = summary_params(results, alpha=alpha)
+        param = summary_params(results, alpha=alpha, use_t=results.use_t)
         info = summary_model(results)
         if xname is not None:
             param.index = xname

statsmodels/regression/linear_model.py

@@ -1106,24 +1106,9 @@ def conf_int(self, alpha=.05, cols=None):
         -----
         The confidence interval is based on Student's t-distribution.
         """
-        bse = self.bse
-        params = self.params
-        # TODO: should be obsolete if super uses use_t
-        if self.use_t:
-            dist = stats.t
-            q = dist.ppf(1 - alpha / 2, self.df_resid)
-        else:
-            dist = stats.norm
-            q = dist.ppf(1 - alpha / 2)
-
-        if cols is None:
-            lower = self.params - q * bse
-            upper = self.params + q * bse
-        else:
-            cols = np.asarray(cols)
-            lower = params[cols] - q * bse[cols]
-            upper = params[cols] + q * bse[cols]
-        return np.asarray(lzip(lower, upper))
+        # keep method for docstring for now
+        ci = super(RegressionResults, self).conf_int(alpha=alpha, cols=cols)
+        return ci
 
 
     @cache_readonly
@@ -1234,13 +1219,6 @@ def f_pvalue(self):
     def bse(self):
         return np.sqrt(np.diag(self.cov_params()))
 
-    @cache_readonly
-    def pvalues(self):
-        # TODO: should be obsolete if super uses use_t
-        if self.use_t:
-            return stats.t.sf(np.abs(self.tvalues), self.df_resid)*2
-        else:
-            return stats.norm.sf(np.abs(self.tvalues))*2
 
     @cache_readonly
     def aic(self):

statsmodels/robust/robust_linear_model.py

@@ -517,7 +517,7 @@ def summary(self, yname=None, xname=None, title=0, alpha=.05,
         smry.add_table_2cols(self, gleft=top_left, gright=top_right, #[],
                           yname=yname, xname=xname, title=title)
         smry.add_table_params(self, yname=yname, xname=xname, alpha=alpha,
-                             use_t=False)
+                             use_t=self.use_t)
 
         #diagnostic table is not used yet
 #        smry.add_table_2cols(self, gleft=diagn_left, gright=diagn_right,