Fix: Incorrect use of partial in TweedieDistribution._rowwise_gradient_hessian

[bili2002]

This PR addresses issue #888 by fixing the incorrect use of partial in the TweedieDistribution._rowwise_gradient_hessian function.

Checklist

• Added a CHANGELOG.rst entry

[stanmart]

Thank you for finding fixing this! I believe there is a similar mistake here:

glum/src/glum/_distribution.py

Line 706 in 9c0a221

f = partial(inv_gaussian_log_eta_mu_deviance, p=self.power)

Could you please address it too so that they are in the same PR?

Also, I think it'd be worth adding a test case that would have spotted these bugs (maybe here?), but I'm also o kay with doing it later.

[stanmart]

Hm, it might be a bit more involved than that. Is it possible that inv_gaussian_log_rowwise_gradient_hessian is incorrect? Shouldn't it be

@@ -276,12 +276,12 @@ def inv_gaussian_log_rowwise_gradient_hessian(
         inv_mu = 1 / mu[i]
         inv_mu2 = inv_mu ** 2
 
-        gradient_rows_out[i] = 2 * weights[i] * (inv_mu - y[i] * inv_mu2)
-        hessian_rows_out[i] = 2 * weights[i] * (inv_mu - 2 * y[i] * inv_mu2)
+        gradient_rows_out[i] = weights[i] * (y[i] * inv_mu2 - inv_mu)
+        hessian_rows_out[i] = weights[i] * (2 * y[i] * inv_mu2 - inv_mu)

instead?

[stanmart]

Thank you :)