Tsunami Warnings: Type I and Type II Errors
Last month I posted on Type I vs. Type II errors, with reference to punitive damages. Now Stephen Karlson at Cold Springs Shops has an excellent piece relating those concepts to tsunami warning systems (and implicitly to the strength of one's Bayesian priors):
We have a classic inference problem. Suppose the null hypothesis is that the earthquake has generated a dangerous tsunami. The government still does not know what its magnitude will be. The beachgoers have no idea how accurate the government's projection is, if one is issued. If you reject the null hypothesis, and it is true, a Type I error, do you drown, or do you see three brief increases in the local surf?
Would your answer be any different if your beach front had a history of tsunamis?
Under what circumstances would you make additional investments in tsunami warning systems?
And people say statistics is dry. I commend this primer on the various sorts of errors in inference researchers may be subject to.
And don't forget to give to some relief programme.
<< Home