R/anosim.R

In vegan: Community Ecology Package

`anosim` <-
    function (x, grouping, permutations = 999,
              distance = "bray", strata = NULL, parallel = getOption("mc.cores"))
{
    EPS <- sqrt(.Machine$double.eps)
    if (!inherits(x, "dist")) { # x is not "dist": try to change it
        if ((is.matrix(x) || is.data.frame(x)) &&
            isSymmetric(unname(as.matrix(x)))) {
            x <- as.dist(x)
            attr(x, "method") <- "user supplied square matrix"
        }
        else
            x <- vegdist(x, method = distance)
    }
    if (any(x < -sqrt(.Machine$double.eps)))
        warning("some dissimilarities are negative - is this intentional?")
    sol <- c(call = match.call())
    grouping <- as.factor(grouping)
    ## check that dims match
    if (length(grouping) != attr(x, "Size"))
        stop(
            gettextf("dissimilarities have %d observations, but grouping has %d",
                     attr(x, "Size"), length(grouping)))
    if (length(levels(grouping)) < 2)
        stop("there should be more than one class level")
    matched <- function(irow, icol, grouping) {
        grouping[irow] == grouping[icol]
    }
    x.rank <- rank(x)
    N <- attr(x, "Size")
    div <- length(x)/2
    irow <- as.vector(as.dist(row(matrix(nrow = N, ncol = N))))
    icol <- as.vector(as.dist(col(matrix(nrow = N, ncol = N))))
    within <- matched(irow, icol, grouping)
    ## check that there is replication
    if (!any(within))
        stop("there should be replicates within groups")
    aver <- tapply(x.rank, within, mean)
    statistic <- -diff(aver)/div
    cl.vec <- rep("Between", length(x))
    take <- as.numeric(irow[within])
    cl.vec[within] <- levels(grouping)[grouping[take]]
    cl.vec <- factor(cl.vec, levels = c("Between", levels(grouping)))
    ptest <- function(take, ...) {
        cl.perm <- grouping[take]
        tmp.within <- matched(irow, icol, cl.perm)
        tmp.ave <- tapply(x.rank, tmp.within, mean)
        -diff(tmp.ave)/div
    }
    permat <- getPermuteMatrix(permutations, N, strata = strata)
    if (ncol(permat) != N)
        stop(gettextf("'permutations' have %d columns, but data have %d rows",
                      ncol(permat), N))
    permutations <- nrow(permat)

    if (permutations) {
        ## Parallel processing
        if (is.null(parallel))
            parallel <- 1
        hasClus <- inherits(parallel, "cluster")
        if (hasClus || parallel > 1) {
            if(.Platform$OS.type == "unix" && !hasClus) {
                perm <- unlist(mclapply(1:permutations,
                                                  function(i, ...)
                                                  ptest(permat[i,]),
                                                  mc.cores = parallel))
            } else {
                if (!hasClus) {
                    parallel <- makeCluster(parallel)
                }
                perm <- parRapply(parallel, permat, ptest)
                if (!hasClus)
                    stopCluster(parallel)
            }
        } else {
            perm <- sapply(1:permutations, function(i) ptest(permat[i,]))
        }
        p.val <- (1 + sum(perm >= statistic - EPS))/(1 + permutations)
    } else { # no permutations
        p.val <- perm <- NA
    }
    sol$signif <- p.val
    sol$perm <- perm
    sol$permutations <- permutations
    sol$statistic <- as.numeric(statistic)
    sol$class.vec <- cl.vec
    sol$dis.rank <- x.rank
    sol$dissimilarity <- attr(x, "method")
    sol$control <- attr(permat, "control")
    class(sol) <- "anosim"
    sol
}