perturbations: Generate Random Variates from a Perturbed RVG

In rvgtest: Tools for Analyzing Non-Uniform Pseudo-Random Variate Generators

Description

These functions perturb the distribution of a given random variate generator. Thus it allows to investigate the power of tests for random variate generators.

Usage

pertadd(n, rdist = rnorm, ..., min = 0, max = 1, p = 0.001)
pertsub(n, rdist = rnorm, ..., min = 0, max = 1, p = 0.001)

Arguments

n	sample size.
rdist	given RVG to be perturbed.
...	parameters for given random variate generator.
min, max	left and right boundary of perturbed domain.
p	strength of perturbation.

Details

pertadd generates random variates from a mixture of rdist and a uniform distribution on the interval (min,max). The uniform distribution is chosen with probability p.

pertsub generates random variates from the rdist but rejects all points in the interval (min,max) with probability p.

By varying the width of uniform distribution (min,max) and probability of error p, different levels of “artificial” error can be introduced. Thus it allows to investigate the power of tests for random variate generators.

Value

A vector of random variates from the perturbed distribution is returned.

Author(s)

Sougata Chaudhuri [e-mail], Josef Leydold [e-mail]

Examples

## Generating random sample with default settings
x <- pertadd(n=1000)
y <- pertsub(n=1000)

## Generating random sample, with parent distribution as exponential.
x <- pertadd(n=1000, rdist=rexp, rate=2, min=2, max=3, p=0.005)
y <- pertsub(n=1000, rdist=rexp, rate=2, min=2, max=3, p=0.005)

## Conducting chi-square test on random variates generated from function
## pertsub with parent distribution as exponential.
ft <- rvgt.ftable(n=1e4,rep=10,
                  rdist=function(n){pertsub(n,rdist=rexp,p=0.1)},
                  qdist=qexp,breaks=1001)
plot(ft)
rvgt.chisq(ft)

## Conducting chi-square test on random variates generated from function
## pertadd with parent distribution as Weibull,shape=1,scale=2.
ft <- rvgt.ftable(n=1e4,rep=10,
                  rdist=function(n,...){pertadd(n,rdist=rweibull,...,p=0.05)},
                  qdist=qweibull,shape=1,scale=2)
plot(ft)
rvgt.chisq(ft)

rvgtest documentation built on May 1, 2019, 6:35 p.m.