Mailbag: Jury regression

Re: Jury science

I have followed your blog with inerest for a while now, and I was looking through your twitter stream and saw this, attributed to Kahneman: "if a court case hinges on regression to the mean, the side that has to explain this to the jury will lose." I was immediately intrigued by the idea of coming up with a good explantion for a jury, and here is what I came up with: Its like poker hands - even if daddy has 4 kings and mommy has 4 queens, the kids aren't going to average much better than two pair, kings over queens. This simplifies and glosses over a lot, but its memorable and avoids the dreaded 'eyes glaze over' effect so common with math explanations for laymen. Perhaps it might be useful in teaching.

Thanks! That does seem appealing. I do regression to the mean over the course of a whole lecture, using data that my students measure on themselves to replicate Galton’s work on stature. I wouldn’t envy anyone who had to do it under a time constraint!