[R] Power curve for a wilcoxon test

[Ben Bolker]

Louisa Hay <louisahay <at> msn.com> writes:

> 
> 
> Hi,
> I am trying to write R code to produce a power curve to show 
> how the power of a 
> Wilcoxon-test varies depending on the mean, with data 
> generated from a uniform
> distribution. Any ideas
> how I should go about this?Louisa

  Here's a brute-force approach that
seems to work reasonably well.  Adapt to 
suit yourself ...

wilcox.pval <- function(m2,m1=0,n=20) {
    x1 <- runif(n,m1-0.5,m1+0.5)
    x2 <- runif(n,m2-0.5,m2+0.5)
    wilcox.test(x1,x2)$p.value
}

wilcox.pow <- function(m2,m1=0,n=20, alpha=0.05, N=500) {
    pvals <- replicate(N,wilcox.pval(m2,m1,n))
    sum(pvals<alpha)/N
}

mvec <- seq(0,0.5,by=0.05)
powvec <- sapply(mvec,wilcox.pow)

plot(mvec,powvec)