SSJ
3.2.1
Stochastic Simulation in Java
|
Tools for Simple Monte Carlo and Quasi-Monte Carlo Experiments. More...
Packages | |
package | anova |
Tools to estimate ANOVA components for Monte Carlo models. | |
Classes | |
class | MonteCarloExperiment |
Provides generic tools to perform simple Monte Carlo experiments with a simulation model that implements one of the interfaces. More... | |
interface | MonteCarloModel |
An interface for a simple simulation model for which Monte Carlo (MC) or RQMC experiments are to be performed. More... | |
interface | MonteCarloModelCV |
An MonteCarloModelDouble that also implements a vector of control variates. More... | |
interface | MonteCarloModelDensityKnown |
An interface for a simulation model for which Monte Carlo (MC) and RQMC experiments are to be performed. More... | |
interface | MonteCarloModelDouble |
An interface for a very simple simulation model for which Monte Carlo (MC) and RQMC experiments are to be performed. More... | |
interface | MonteCarloModelDoubleArray |
Similar to MonteCarloModelDouble except that the returned performance is an array of real numbers. More... | |
class | RQMCExperiment |
Provides basic generic tools to perform RQMC experiments with a simulation model that implements the MonteCarloModelDouble interface. More... | |
class | RQMCExperimentSeries |
This class offers facilities to perform experiments on the convergence of the variance when estimating a mean (expectation) with a series of RQMC point sets usually of the same type, but different sizes \(n\). More... | |
Tools for Simple Monte Carlo and Quasi-Monte Carlo Experiments.
This package offers elementary tools to help perform Monte Carlo and quasi-Monte Carlo experiments with simple stochastic simulation (Monte Carlo) models.
To use these tools, the simulation model must be represented in a class that implements the MonteCarloModel interface. This interface requires in particular a method simulate to simulate the model and a method getPerformance to recover the performance measure(s) (or data) returned by the simulation. This `‘performance’' can be any type of object. In the simplest case, it is a real number; this is what is assumed in the MonteCarloModelDouble class. In the class MonteCarloModelDoubleArray, the returned performance is assumed to be a double[] array, so a vector of several measures can be returned for each simulation run. In MonteCarloModelCV, it is assumed that the simulation produces a real-valued performance, plus a vector of zero-mean control variates.
The class MonteCarloExperiment provides methods to perform a simulation experiment that simulates the model for #n independent runs, then recover the data in statistical collectors, compute confidence intervals, make histograms, scatter plots, etc.
The class RQMCExperiment provides tools to perform similar experiments, but with randomized quasi-Monte Carlo (RQMC) instead. In an RQMC experiment, one simulates the system #n times, using an RQMC point set of size #n, and this is repeated #r times by randomizing the point set independently for each of the #r replications.
RQMCExperimentSeries can perform RQMC experiments for a series of several values of #n, for example all powers of 2 in some interval, in order to study the behavior of RQMC estimators as a function of #n. For example, one might be interested in estimating how the variance (and in some cases the bias) converges as a function of #n, and perhaps plot it in log-log scale.
The following examples show how to use this package. (To do... )