"evaluate.profiles" <-
function(npl, group.labels, funstat=max, times=1000){
        ## npl: matrix of m pedigrees by n positions, from Genehunter
        ## group.labels: labels for groups
        ## funstat: funstat=max takes the maximum over all profiles.
        ## times: permute labels this number of times
        ## i.e., take sample of size times from permutation distribution.
        funprof <- function(m, group, fun=funstat, profiles=FALSE){
            gp.subscripts <- split(1:m, group)
            gpprofs <- sapply(gp.subscripts,
                              function(x, npl)apply(npl[x,], 2, sum),
                              npl=npl)
            gpprofs <- sweep(gpprofs,2, sqrt(num.in.gps), "/")
            if(profiles) gpprofs else fun(gpprofs)
        }
        m <- dim(npl)[1]
        if(m != length(group.labels))
            stop("# of group labels must equal # of rows of npl")
        num.in.gps <- table(group.labels)
        ngps <- length(tab)
        group.profiles <- funprof(m=m, group=group.labels, profiles=TRUE)
        samp.stat <- funstat(group.profiles)
        permstats <- numeric(times)
        for(i in 1:times){
            newgp <- group.labels[sample(m)]
            permstats[i] <- funprof(m=m, group=newgp, profiles=FALSE)
        }
        p <- 1-sum(samp.stat>permstats)/times
        list(p=p, profiles=group.profiles)
    }