When estimating a parameter in log, these were not scaled appropriately and...

When estimating a parameter in log, these were not scaled appropriately and taken in log when saving the related PDF

CHANGELOG.md

@@ -7,7 +7,7 @@
 ### Changed
 ### Fixed
 - To estimate parameters in log, pcigale determines which variables end with the "_log" string and removed it to compute the models. However in some circumstances, it was overzealous. This has been fixed. (Médéric Boquien)
-- When estimating a parameter in log, these were not scaled appropriately and taken in log when saving the related χ². (Médéric Boquien)
+- When estimating a parameter in log, these were not scaled appropriately and taken in log when saving the related χ² and PDF. (Médéric Boquien)
 
 ### Optimised
 - Prior to version 0.7.0, we needed to maintain the list of redshifts for all the computed models. Past 0.7.0 we just infer the redshift from a list unique redshifts. This means that we can now discard the list of redshifts for all the models and only keep the list of unique redshifts. This saves ~8 MB of memory for every 10⁶ models. the models should be computed slightly faster but it is in the measurement noise. (Médéric Boquien)

pcigale/analysis_modules/pdf_analysis/utils.py

@@ -32,7 +32,7 @@ def save_best_sed(obsid, sed, norm):
     sed.to_votable(OUT_DIR + "{}_best_model.xml".format(obsid), mass=norm)
 
 
-def save_pdf(obsid, name, model_variable, likelihood):
+def _save_pdf(obsid, name, model_variable, likelihood):
     """Compute and save the PDF to a FITS file
 
     We estimate the probability density functions (PDF) of the parameter from
@@ -83,6 +83,39 @@ def save_pdf(obsid, name, model_variable, likelihood):
         table.write(OUT_DIR + "{}_{}_pdf.fits".format(obsid, name))
 
 
+def save_pdf(obsid, names, mass_proportional, model_variables, scaling,
+             likelihood, wlikely):
+    """Compute and save the PDF of analysed variables
+
+    Parameters
+    ----------
+    obsid: string
+        Name of the object. Used to prepend the output file name
+    names: list of strings
+        Analysed variables names
+    model_variables: array
+        Values of the model variables
+    likelihood: 1D array
+        Likelihood of the "likely" models
+
+    """
+    for i, name in enumerate(names):
+        if name.endswith('_log'):
+            if name[:-4] in mass_proportional:
+                model_variable = np.log10(model_variables[:, i][wlikely] *
+                                          scaling[wlikely])
+            else:
+                model_variable = np.log10(model_variables[:, i][wlikely])
+        else:
+            if name in mass_proportional:
+                model_variable = (model_variables[:, i][wlikely] *
+                                  scaling[wlikely])
+            else:
+                model_variable = model_variables[:, i][wlikely]
+
+        _save_pdf(obsid, name, model_variable, likelihood)
+
+
 def _save_chi2(obsid, name, model_variable, chi2):
     """Save the best reduced χ² versus an analysed variable
 

pcigale/analysis_modules/pdf_analysis/workers.py

@@ -291,14 +291,9 @@ def analysis(idx, obs):
                       sed.mass_proportional_info, gbl_model_variables[wz, :],
                       scaling, chi2 / (nobs - 1))
         if gbl_save['pdf']:
-            for i, variable in enumerate(gbl_analysed_variables):
-                if variable in sed.mass_proportional_info:
-                    save_pdf(obs['id'], variable,
-                             gbl_model_variables[wz, i][wlikely] *
-                             scaling[wlikely], likelihood)
-                else:
-                    save_pdf(obs['id'], variable,
-                             gbl_model_variables[wz, i][wlikely], likelihood)
+            save_pdf(obs['id'], gbl_analysed_variables,
+                     sed.mass_proportional_info, gbl_model_variables[wz, :],
+                     scaling, likelihood, wlikely)
 
     with gbl_n_computed.get_lock():
         gbl_n_computed.value += 1