A homework question in someone’s 11th grade statistics class

Andrew

Of the four options given above, I think option A is the best, then D is the second best, but the best choice is not completely clear. The trouble with option A is that the norespondents might not respond even after encouragement. The trouble with option D is that the nonrespondents can differ from the respondents. Options B and E are fine too, as they both involve gathering more data, but presumably the new data will still have noresponse issues.

Usual good practice would be to gather available data on the respondents and use this information to adjust the sample to the population. We discuss this in chapter 17 of Regression and Other Stories. This would go beyond what is taught in high school, though. I actually don’t like this sort of message that says that nonrandom samples are bad but without a discussion of how to fix the problems. I hope someday to design a high-school-level statistics course, but I’m not there yet.

You may also like