> lst = list() > m = 250; n = 10 > for(i in 1:m) { + lst$exp[i] = t.test(rexp(n), mu=1, df=n-1)$p.value + lst$unif[i] = t.test(runif(n), mu=.5, df=n-1)$p.value + lst$t4[i] = t.test(rt(n, df=4), mu=0, df=n-1)$p.value + } > sapply(lst, function(x) sum(x<0.05)/length(x)) exp unif t4 0.088 0.052 0.044When the population distribution is skewed, the sampling distribution of T may differ a lot from the t-distribution.