Plotting Upgrade #137

Activity

1. ihincks

   added

   enhancement

   on Aug 1, 2017
2. ihincks commented on Aug 18, 2017

   

   ihincks

   on Aug 18, 2017

   Last edited by ihincks

   CollaboratorAuthor

   More actions

   The seaborn library works most conveniently when the data is presented using pandas data types, like DataFrame. Perhaps a modest first step would just be to add a dataframe method to ParticleDistribution, or perhaps to SMCUpdater (which has access to modelparam_names). Something along the lines of (untested):

   @property
   def dataframe(self):
       # we want uniform particle weights
       dist = self if self.just_resampled else self.resampler(self)
       return pandas.DataFrame(data={
           name: dist.particle_locations[:,idx] 
           for idx, name in enumerate(self.model.modelparam_names)
       })
   

   We don't have to make pandas a dependency if we don't want to.

   EDIT: KDEs can be expensive, so maybe it should be a method with some sort of slice option.