Add cash_sum_cython and cstat_sum_cython reference tests

gammapy/cube/fit.py

@@ -165,6 +165,8 @@ def likelihood(self, parameters, mask=None):
         """
         counts, npred = self._counts_data, self.npred().data
 
+        #TODO: add mask handling to _stat_sum, so that the temp copy
+        # created by the fancy indexing is avoided
         if self.mask is None and mask is None:
             stat = self._stat_sum(counts.ravel(), npred.ravel())
         elif self.mask is None:

gammapy/stats/tests/test_fit_statistics.py

@@ -105,18 +105,32 @@ def test_wstat(test_data, reference_values):
     assert_allclose(statsvec, reference_values["wstat"])
 
 
-@requires_dependency("sherpa")
 def test_cash(test_data, reference_values):
     statsvec = stats.cash(n_on=test_data["n_on"], mu_on=test_data["mu_sig"])
     assert_allclose(statsvec, reference_values["cash"])
 
 
-@requires_dependency("sherpa")
 def test_cstat(test_data, reference_values):
     statsvec = stats.cstat(n_on=test_data["n_on"], mu_on=test_data["mu_sig"])
     assert_allclose(statsvec, reference_values["cstat"])
 
 
+def test_cash_sum_cython(test_data):
+    counts = np.array(test_data["n_on"], dtype=float)
+    npred = np.array(test_data["mu_sig"], dtype=float)
+    stat = stats.cash_sum_cython(counts=counts, npred=npred)
+    ref = stats.cash(counts, npred).sum()
+    assert_allclose(stat, ref)
+
+
+def test_ctstat_sum_cython(test_data):
+    counts = np.array(test_data["n_on"], dtype=float)
+    npred = np.array(test_data["mu_sig"], dtype=float)
+    stat = stats.cstat_sum_cython(counts=counts, npred=npred)
+    ref = stats.cstat(counts, npred).sum()
+    assert_allclose(stat, ref)
+
+
 def test_wstat_corner_cases():
     """test WSTAT formulae for corner cases"""
     n_on = 0