More on the evils of statistical tests


Check out the two posts by Galit Shmueli over at Bzst on hypothesis tests: one on the value of p-values and another on one-sided tests.

She says “Shockingly enough, people seem to really want to use p-values, even if they don’t understand them.” That mirrors my experience too.  Confidence intervals are much more useful because they provide a measure of the size of an effect, rather than testing if it is equal to some prespecified value.

comments powered by Disqus