Under the Errata forum in a thread I called Second Print Run Corrections , one poster replied that he believed I incorrectly applied the term confidence interval in the book. I discuss several errors in that post in a reply in that thread. But it introduces another point of confusion apparently held by some about the difference between Bayesian vs. non-Bayesian methods in statistics and the epistemicologicaly philosophy debate of the frequentist vs. the subjectivist. I addressed it in another thread called Bayesian vs. Frequentist in this In the Clouds forum topic.
Bayesian vs. Frequentist?
by Douglas Hubbard | Bayesian vs. Frequentist?, Errata, How To Measure Anything Blogs, News
Read an interesting article today that argues quite effectively that humans are natural frequentists. I believe that the author is a staunch supporter of the Bayesian philosophy though.
What do you think?
Thanks for your contribution.
Yes, you are correct that Gigerenzer is actually a Bayesian on the point of the way to interpret the meaning of probabilities. His observations do, however, tell us some teaching methods that measurably improve the ellicitation of subjective probabilities.
But there are some other findings that point in the opposite direction of Gigerenzer’s findings. In my book I cite work by Grether who shows that when people are given very specific situations with prior probabilities and new information, then tend to update their knowledge in a way that tis not inconsistent with Bayes (even though they are not aware of this theory). Kahnemann and others have shown, however, that in other situations people tend to ignore prior probabilities. So I think it’s more complicated than just whether we are instinctively frequentist or Bayesian. Perhaps Gigenrenzer was simply showing that less abstract concepts are easier to estimate while Grether showed that when people are given new information together with prior information, they are not entirely irrational.
But both do have immediate practical impacts for decision makers. I use Gigerenzer’s methods among others when I train people to provide more realistic probabilities and I use Grether’s findings as the basis for the validity of what I call the “instinctive Bayesian” method of measurement in the book. But there is definitely a lot more to research in this area.