Navigation

Global	Documentation
Home	Manual
Readme	Glossary
Notes	Design
Examples	Classes
	Trouble
	Style

class Sample( object )	Source

Sample is weighted random draw from a Posterior distribution as provided by a Sampler

Each Sample maintains 5 attributes

Attributes

• id : int
       identification number
• parent : int
       id of the parent (-1 for Adam/Eve)
• model : Model
       the model being used
• logL : float
       log Likelihood = log Prob( data | params )
• logW : float
       log Weights of the log of the weight of the sample.
       The weight is the relative contribution to the evidence integral.
       logW = logL + log( width )
       The logZ, the evidence, equals the log of the sum of the contributions.
       logZ = log( sum( exp( logW ) ) )
• parameters : array_like
       parameters (of the model)
• nuisance : array_like (optional)
       nuisance parameters (of the problem)
• hyper : array_like (optional)
       list of hyper parameters (of the error distribution)
• fitIndex : array_like or None
       list of allpars to be fitted.
• allpars : array_like (read only)
       list of parameters, nuisance parameters and hyperparameters

Author Do Kester

Sample( id, parent, start, model, parameters=None, fitIndex=None, copy=None )

Constructor.

Parameters

• id : int
       id of the sample
• parent : int
       id of the parent (-1 for Adam/Eve)
• start : int
       iteration in which the walker was constructed
• model : Model
       the model being used. Parameters are copied from this model.
• parameters : array_like
       list of model parameters
• fitIndex : array_like
       list of indices in allpars that need fitting
• copy : Sample
       the sample to be copied

copy( )

Copy.