diff --git a/statsmodels/genmod/tests/test_glm_weights.py b/statsmodels/genmod/tests/test_glm_weights.py
index 9db40a2a484..2ad93ea2476 100644
--- a/statsmodels/genmod/tests/test_glm_weights.py
+++ b/statsmodels/genmod/tests/test_glm_weights.py
@@ -244,11 +244,9 @@ def setup_class(cls):
         nobs = len(cpunish_data.endog)
         aweights = fweights / wsum * nobs
 
-        # This is really close when corr_fact = (wsum - 1.) / wsum, but to
-        # avoid having loosen precision of the assert_allclose, I'm doing this
-        # manually. Its *possible* lowering the IRLS convergence criterion
-        # in stata and here will make this less sketchy.
-        cls.corr_fact = np.sqrt((wsum - 1.) / wsum) * 0.98518473599905609
+        # Note: corr_fact here matches one of the two ways it can be defined
+        #  in test_sandwich_cov.
+        cls.corr_fact = np.sqrt((nobs - 1.) / nobs)
         mod = GLM(cpunish_data.endog, cpunish_data.exog,
                   family=sm.families.Poisson(),
                   var_weights=aweights)