Fix NegativeBinomialDistribution.theta type check to accept np.number (#977)

The theta setter only accepted (int, float), but internally stores
values as np.float32. This caused dist.theta = dist.theta to raise
a TypeError. Add np.number to the type check, matching the fix
already applied to TweedieDistribution.power.

Fixes #631

Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: danielmeint

CHANGELOG.rst

@@ -13,6 +13,7 @@ Changelog
 **Bug fix:**
 
 - Fixed ``deviance_path_`` in :class:`~glum.GeneralizedLinearRegressorCV` being scaled down by a factor of ``n_folds`` because test fold weights were not normalized to sum to 1.
+- Fixed :class:`~glum.NegativeBinomialDistribution` ``theta`` setter rejecting ``np.number`` types, causing ``dist.theta = dist.theta`` to raise a ``TypeError``.
 
 **Other changes:**
 

src/glum/_distribution.py

@@ -1390,7 +1390,7 @@ def theta(self):
 
     @theta.setter
     def theta(self, theta):
-        if not isinstance(theta, (int, float)):
+        if not isinstance(theta, (int, float, np.number)):
             raise TypeError(f"Theta must be numeric; got {theta}.")
         if not theta > 0:
             raise ValueError(f"Theta must be strictly positive; got was {theta}.")

tests/glm/test_distribution.py

@@ -98,6 +98,17 @@ def test_negative_binomial_distribution_alpha():
         dist.theta = 1j
 
 
+def test_distribution_parameter_reassignment():
+    """Test that distribution parameters can be reassigned to themselves."""
+    dist = NegativeBinomialDistribution(theta=1.5)
+    dist.theta = dist.theta
+    assert dist.theta == 1.5
+
+    dist2 = TweedieDistribution(power=1.5)
+    dist2.power = dist2.power
+    assert dist2.power == 1.5
+
+
 def test_negative_binomial_distribution_parsing():
     dist = get_family("negative.binomial")