Merge pull request statsmodels#8907 from josef-pkt/bug_mnlogit_wald_tests

BUG: mnlogit wald tests, ravel, string cov_names

Author: josef-pkt

statsmodels/base/model.py

@@ -1639,7 +1639,11 @@ def t_test(self, r_matrix, cov_p=None, use_t=None):
         """
         from patsy import DesignInfo
         use_t = bool_like(use_t, "use_t", strict=True, optional=True)
-        names = self.model.data.cov_names
+        if self.params.ndim == 2:
+            names = ['y{}_{}'.format(i[0], i[1])
+                     for i in self.model.data.cov_names]
+        else:
+            names = self.model.data.cov_names
         LC = DesignInfo(names).linear_constraint(r_matrix)
         r_matrix, q_matrix = LC.coefs, LC.constants
         num_ttests = r_matrix.shape[0]
@@ -1649,7 +1653,7 @@ def t_test(self, r_matrix, cov_p=None, use_t=None):
                 not hasattr(self, 'cov_params_default')):
             raise ValueError('Need covariance of parameters for computing '
                              'T statistics')
-        params = self.params.ravel()
+        params = self.params.ravel(order="F")
         if num_params != params.shape[0]:
             raise ValueError('r_matrix and params are not aligned')
         if q_matrix is None:
@@ -1853,8 +1857,12 @@ def wald_test(self, r_matrix, cov_p=None, invcov=None,
             use_f = (hasattr(self, 'use_t') and self.use_t)
 
         from patsy import DesignInfo
-        names = self.model.data.cov_names
-        params = self.params.ravel()
+        if self.params.ndim == 2:
+            names = ['y{}_{}'.format(i[0], i[1])
+                     for i in self.model.data.cov_names]
+        else:
+            names = self.model.data.cov_names
+        params = self.params.ravel(order="F")
         LC = DesignInfo(names).linear_constraint(r_matrix)
         r_matrix, q_matrix = LC.coefs, LC.constants
 

statsmodels/discrete/tests/test_discrete.py

@@ -2586,16 +2586,15 @@ def test_cov_confint_pandas():
     assert isinstance(ci.index, pd.MultiIndex)
 
 
-def test_t_test():
+def test_mlogit_t_test():
     # GH669, check t_test works in multivariate model
-    data = load_anes96()
+    data = sm.datasets.anes96.load()
     exog = sm.add_constant(data.exog, prepend=False)
     res1 = sm.MNLogit(data.endog, exog).fit(disp=0)
     r = np.ones(res1.cov_params().shape[0])
     t1 = res1.t_test(r)
     f1 = res1.f_test(r)
 
-    data = sm.datasets.anes96.load()
     exog = sm.add_constant(data.exog, prepend=False)
     endog, exog = np.asarray(data.endog), np.asarray(exog)
     res2 = sm.MNLogit(endog, exog).fit(disp=0)
@@ -2604,3 +2603,20 @@ def test_t_test():
 
     assert_allclose(t1.effect, t2.effect)
     assert_allclose(f1.statistic, f2.statistic)
+
+    tt = res1.t_test(np.eye(np.size(res2.params)))
+    assert_allclose(tt.tvalue.reshape(6,6, order="F"), res1.tvalues.to_numpy())
+    tt = res2.t_test(np.eye(np.size(res2.params)))
+    assert_allclose(tt.tvalue.reshape(6,6, order="F"), res2.tvalues)
+
+    wt = res1.wald_test(np.eye(np.size(res2.params))[0], scalar=True)
+    assert_allclose(wt.pvalue, res1.pvalues.to_numpy()[0, 0])
+
+
+    tt = res1.t_test("y1_logpopul")
+    wt = res1.wald_test("y1_logpopul", scalar=True)
+    assert_allclose(tt.pvalue, wt.pvalue)
+
+    wt = res1.wald_test("y1_logpopul, y2_logpopul", scalar=True)
+    # regression test
+    assert_allclose(wt.statistic, 5.68660562, rtol=1e-8)