Donald Knuth on Stratified Random Sampling
In Chapter 2 of Things a Computer Scientist Rarely Talks About, Knuth said about his use of stratified random sampling in Bible study:
We could have rolled dice. True randomization clearly leads to a better sample than the results of a fixed deterministic like chapter 3, verse 16. On the other hand, there is no reason to think there is anything unusual about chapter 3, verse 16 except in the Book of John. And I didn't like the idea of rolling dice, for several reasons. One reason was that we would have to roll the dice in advance if people were going to prepare for the class. But then if a person missed a class they wouldn't know what to do for next time. The other reason is that when you roll dice there's a temptation to cheat. You get a bad roll and you say, “Well I didn't really mean that … the dice slipped, or bumped into the edge. Let's try again.” Thus my 3:16 rule actually couldn't be rigged.