broadwick.abc

Class ApproxBayesianComp

java.lang.Object

broadwick.abc.ApproxBayesianComp

public class ApproxBayesianComp
extends Object

This class runs an approximate Bayesian computation, sampling from the posterior.

Field Summary

Fields

Modifier and Type	Field and Description
private AbcController	controller
private AbcDistance	distance
private double	epsilon
private AbcModel	model
private int	numSamplesTaken
private AbcNamedQuantity	observedData
private Map<String,LinkedList<Double>>	posteriors
private AbcPriorsSampler	priors

Constructor Summary

Constructors

Constructor and Description
ApproxBayesianComp(AbcNamedQuantity observedData, AbcModel model, AbcPriorsSampler priors, double sensitivity) Create an ABC instance.

Method Summary

Modifier and Type	Method and Description
void	run() Run the ABC algorithm.
private void	save(AbcNamedQuantity prior) Save the sample taken from the prior as it meets the criteria set for being a posterior sample.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

controller

private AbcController controller

distance

private AbcDistance distance

model

private AbcModel model

priors

private AbcPriorsSampler priors

posteriors

private Map<String,LinkedList<Double>> posteriors

observedData

private AbcNamedQuantity observedData

epsilon

private double epsilon

numSamplesTaken

private int numSamplesTaken

Constructor Detail

ApproxBayesianComp

public ApproxBayesianComp(AbcNamedQuantity observedData,
                          AbcModel model,
                          AbcPriorsSampler priors,
                          double sensitivity)

Create an ABC instance. The created instance will be ready to run a calculation using the default controller and distance measure, these will need to be set if custom ones are required.

Parameters:

observedData - the observed data from which the distance will be calculated.

model - the model that generates the data.

priors - the sampler objec that samples from the priors.

sensitivity - the distance cutoff, distance greater than this value will be ignored.

Method Detail

run

public final void run()

Run the ABC algorithm.

save

private void save(AbcNamedQuantity prior)

Save the sample taken from the prior as it meets the criteria set for being a posterior sample.

Parameters:

prior - the sampled values from the prior.