The article clarifies two points about statistical inference. First, when doing 20 hypothesis tests, the chance of making a type I error is around 64%. Second, large sample sizes improve accuracy of estimates mainly due to more data, not because they approach the population size.