R2sample (version 4.1.0)

twosample_power: Power estimation for two-sample methods

Description

Find the power of various two sample tests using Rcpp and parallel computing.

Usage

twosample_power(
  f,
  ...,
  TS,
  TSextra,
  With.p.value = FALSE,
  alpha = 0.05,
  B = 1000,
  nbins = c(50, 10),
  minexpcount = 5,
  UseLargeSample,
  samplingmethod = "Binomial",
  rnull,
  SuppressMessages = FALSE,
  maxProcessor
)

Value

A numeric vector of power values.

Arguments

f

function to generate a list with data sets x, y and (optional) vals, weights

...

additional arguments passed to f, up to 2

TS

routine to calculate test statistics for non-chi-square tests

TSextra

additional info passed to TS, if necessary

With.p.value

=FALSE does user supplied routine return p values?

alpha

=0.05, the level of the hypothesis test

B

=1000, number of simulation runs.

nbins

=c(50,10), number of bins for chi large and chi small.

minexpcount

=5 minimum required count for chi square tests

UseLargeSample

should p values be found via large sample theory if n,m>10000?

samplingmethod

="Binomial" or independence in discrete data case

rnull

a function that generates data from a model, possibly with parameter estimation.

SuppressMessages

= FALSE print informative messages?

maxProcessor

maximum number of cores to use. If maxProcessor=1 no parallel computing is used.

Details

For details consult vignette("R2sample","R2sample")

This routine runs a number of different two-sample tests for univariate data, either discrete or continuous. The user can also provide their own test method.

Examples

 # Power of standard normal vs. normal with mean mu.
 f1=function(mu) list(x=rnorm(25), y=rnorm(25, mu))
 #Power of uniform discrete distribution vs. with different probabilities.
 twosample_power(f1, mu=c(0,2), B=100, maxProcessor = 1)
 f2=function(n, p) list(x=table(sample(1:5, size=1000, replace=TRUE)), 
       y=table(sample(1:5, size=n, replace=TRUE, 
       prob=c(1, 1, 1, 1, p))), vals=1:5)
 twosample_power(f2, n=c(1000, 2000), p=c(1, 1.5), B=100, maxProcessor = 1)
 # Compare power of a new test with those in package:
 myTS=function(x,y) {z=c(mean(x)-mean(y),sd(x)-sd(y));names(z)=c("M","S");z}
 cbind(twosample_power(f1, mu=c(0,2), TS=myTS,B=100, maxProcessor = 1),
       twosample_power(f1, mu=c(0,2), B=100, maxProcessor = 1))
 # Power estimation if routine returns a p value
 myTS2=function(x, y) {out=ks.test(x,y)$p.value; names(out)="KSp"; out}      
 twosample_power(f1, c(0,1), TS=myTS2, With.p.value = TRUE,  B=100)