Change start_params to None in call to GLM.fit_constrained used to ge…

…t automatic priors (#265)

* Change start_params to None in GLM.fit_constrained

* Black format

* Omit start_params. It is None by default

* Update changelog

Changelog.md

@@ -18,6 +18,7 @@
 * Include missing files for sdist (#204)
 * Fixed if-else comparison that prevented HalfTStudent prior from being used (#205)
 * Sidestep plotting flat priors in `plot_priors()` (#258)
+* GLM.fit_constrained in automatic priors now uses start_params = None (#265)
 
 ### Documentation
 * Update example notebooks (#232)

bambi/priors.py

@@ -248,9 +248,7 @@ def _get_slope_stats(self, exog, predictor, sigma_corr, full_mod=None, points=4)
             null = [
                 GLM(
                     endog=self.model.y.data, exog=exog, family=self.model.family.smfamily()
-                ).fit_constrained(
-                    str(exog.columns[i]) + "=" + str(val), start_params=full_mod.params.values
-                )
+                ).fit_constrained(str(exog.columns[i]) + "=" + str(val))
                 for val in values[:-1]
             ]
             null = np.append(null, full_mod)