Nested sampling with MultiNest

docs/faq.rst

@@ -27,10 +27,10 @@ imaged planets. A detailed description of how τ is related to other quantities
 
 Our goal with the default prior is to have an isotropic distribution of the orbital plane.
 To obtain this, we use inclination and position angle of the ascending node (PAN) to
-define the orbital plane. They actually coorespond to the two angles in a spherical coordinate system:
-inclinaion is the polar angle and PAN is the azimuthal angle. Becuase of this choice of coordinates,
+define the orbital plane. They actually correspond to the two angles in a spherical coordinate system:
+inclination is the polar angle and PAN is the azimuthal angle. Because of this choice of coordinates,
 there are less orbital configurations near the poles (when inclination is near 0 or 180), so we use
-a sine prior to downweigh those areas to give us an isotropic distribution. 
+a sine prior to downweight those areas to give us an isotropic distribution. 
 A more detailed discussion of this is here:
 
 .. toctree::

orbitize/driver.py

@@ -16,7 +16,9 @@ class Driver(object):
         input_data: Either a relative path to data file or astropy.table.Table object
             in the orbitize format. See ``orbitize.read_input``
         sampler_str (str): algorithm to use for orbit computation. "MCMC" for
-            Markov Chain Monte Carlo, "OFTI" for Orbits for the Impatient
+            Markov Chain Monte Carlo, "OFTI" for Orbits for the Impatient,
+            "NestedSampler" for nested sampling with Dynesty, or "MultiNest"
+            for nested sampling with (Py)MultiNest.
         num_secondary_bodies (int): number of secondary bodies in the system.
             Should be at least 1.
         stellar_or_system_mass (float): mass of the primary star (if fitting for

orbitize/results.py

@@ -45,6 +45,8 @@ def __init__(
         self.lnlike = lnlike
         self.curr_pos = curr_pos
         self.version_number = version_number
+        self.ln_evidence = None
+        self.ln_evidence_err = None
 
         if self.system is not None:
             self.tau_ref_epoch = self.system.tau_ref_epoch
@@ -114,6 +116,12 @@ def save_results(self, filename):
         hf.attrs['sampler_name'] = self.sampler_name
         hf.attrs['version_number'] = self.version_number
 
+        if self.ln_evidence is not None:
+            hf.attrs['ln_evidence'] = self.ln_evidence
+
+        if self.ln_evidence_err is not None:
+            hf.attrs['ln_evidence_err'] = self.ln_evidence_err
+
         # Now add post and lnlike from the results object as datasets
         hf.create_dataset('post', data=self.post)
         # hf.create_dataset('data', data=self.data)
@@ -147,19 +155,18 @@ def load_results(self, filename, append=False):
         hf = h5py.File(filename, 'r')  # Opens file for reading
         # Load up each dataset from hdf5 file
         sampler_name = str(hf.attrs['sampler_name'])
-        try:
+        if 'version_number' in hf.attrs:
             version_number = str(hf.attrs['version_number'])
-        except KeyError:
+        else:
             version_number = "<= 1.13"
         post = np.array(hf.get('post'))
         lnlike = np.array(hf.get('lnlike'))
 
-
-        try:
+        if 'num_secondary_bodies' in hf.attrs:
             num_secondary_bodies = int(hf.attrs['num_secondary_bodies'])
-        except KeyError:
+        else:
             # old, has to be single planet fit
-            num_secondary_bodies = 1  
+            num_secondary_bodies = 1
 
         try:
             data_table = table.Table(np.array(hf.get('data')))