Watson: Strange business, eh, Holmes? What do you make of the strangling of Mr. Evers at the Evers mansion. Fortunately, they found blood of the assailant under Mr. Evers' fingernails. The whole family is under suspicion pending the blood tests. Of course, the blood tests may clear the whole lot. What's your guess?
Holmes: My hypothesis is that the blood stains will be type O, the same blood type as Mrs. Evers, the victim's wife.
Watson: I just can't see her having the strength or viciousness to do that!
Holmes: Quite! But what would it prove it the blood stains matched?
Watson: It's obvious. If the blood stains matched, she would have the same blood type as the attacker. Therefore, uh, therefore, what? Holmes, I fail to understand what that would mean.
Holmes: Why are you confused, Watson?
Watson: That would prove she had the same blood type as the attacker, but millions of people in this city have that blood type.
Holmes: Correct, Watson.
Watson: I don't follow, Holmes.
Holmes: I've used the logic of null hypothesis testing. I've stated that I expect no difference between her blood and the blood at the scene. Given our current technology, this hypothesis can never be proven correct. For example, even if both her blood and the blood at the scene were type O, we could not say that her blood was at the crime scene.
Watson: Okay, but why state this "no difference" hypothesis if you can't prove it?
Holmes: Elementary, my dear Watson. What happens if her blood type is O and the assailant's blood type is A? My null hypothesis has been disproven, so we can be sure that Mrs. Evers was not the assailant--which is what I believe. Indeed, Mrs. Evers is our client.
Watson: Why didn't you say that in the first place?
Holmes: And waste this opportunity to instruct you? My dear chap, you know my methods.
To Research Design Home Page