regression for horse problem

regressionhorse.R

set.seed(10)
lm.power <- c()
spd.power <- c()
for (n in range) {
    spd.pvalue <- c()
    lm.pvalue <- c()
    for (j in 1:1E2) {
        mat <- c()
        samplesize <- 1E2
        for (i in 1:samplesize) {
            mat <- rbind(mat,sample(1:n, prob=c(rep(1.2,floor(n / 2)), rep(1, n - floor(n / 2)))))
        }
        distances <- apply(mat, 1, permdist)
        logd <- apply(mat, 1, tolodds)
        # creating log-odds of start
        n1 <- (1:n) / (n+1)
        n2 <- log(n1 / (1 - n1))
        logn <- rep(n2,samplesize)
        # THIS LINE IS THE REGRESSION THING. THE OTHER LINES ARE NOT THAT IMPORTANT
        # calculate the p-value of the linear model test
        lm.pvalue <- c(lm.pvalue,summary(lm(as.vector(logd)~logn))$coef[2,4])
        spd <- (distances - means[n-first+1])/sds[n-first+1]
        average <- mean(spd)
        # calculate the p-value of the permutation test (CLT)
        spd.pvalue <- c(spd.pvalue, pnorm(average*sqrt(length(distances))))
    }
    # calculate the simulated power
    spd.power <- c(spd.power, sum(spd.pvalue < 0.05 | spd.pvalue > 0.95)/length(spd.pvalue))
    lm.power <- c(lm.power, sum(lm.pvalue < 0.05 | lm.pvalue > 0.95)/length(lm.pvalue))
}