Give an example of when a one-sided test on a population mean would make more sense than a two tailed test.
Give an example of the opposite.
In general, why do we say that there is no statistical way to decide whether a test should be run as a one-tailed test or a two-tailed test?