marginals returned now with integral 1, so that plotting works

psignifit/_posterior.py

@@ -242,9 +242,13 @@ def objective(x):
 def marginalize_posterior(parameter_grid: ParameterGrid, posterior_mass: np.ndarray) -> Dict[str, np.ndarray]:
     marginals = dict()
     for i, (param, grid) in enumerate(sorted(parameter_grid.items())):
-        if grid is None:
+        if grid is None or len(grid)==1:
             marginals[param] = None
-        axis = tuple(range(0, i)) + tuple(range(i + 1, len(parameter_grid)))
-        marginals[param] = np.squeeze(posterior_mass.sum(axis))
+        else:
+            axis = tuple(range(0, i)) + tuple(range(i + 1, len(parameter_grid)))
+            # we get first the unnormalized marginal, and then we scale it
+            nmarginal = np.squeeze(posterior_mass.sum(axis))
+            integral = np.trapz(nmarginal, x=grid)
+            marginals[param] = nmarginal / integral
 
     return marginals

psignifit/psignifit.py

@@ -104,9 +104,9 @@ def psignifit(data: np.ndarray, conf: Optional[Configuration] = None,
     return Result(parameter_estimate=fit_dict,
                   configuration=conf,
                   confidence_intervals=intervals_dict,
-                  parameter_values={k: v.tolist() for k, v in grid.items()},
-                  prior_values={param: priors[param](values).tolist() for param, values in grid.items()},
-                  marginal_posterior_values={k: v.tolist() for k, v in marginals.items()},
+                  parameter_values=grid,
+                  prior_values={param: priors[param](values) for param, values in grid.items()},
+                  marginal_posterior_values=marginals,
                   posterior_mass=posteriors,
                   data=data.tolist())