Re:Rules for Choosing Nonfiction Books
by wenc on Friday February 08, 2019 @03:05PM
(#19118133)
That's a good point -- I wish we could see Sparklines-like histograms of reviews.
On a related note, for binary (good/bad) reviews, we can the temper the results by assuming a beta(1,1)-distributed prior and then updating [1]. The expectation (which can thought of as an "adjusted average rating") ends up being:
E[x] = (good + 1)/(good + bad + 2)
This adjusts for situations where the number of reviews is small, in which a good average rating can be misleading.
On a related note, for binary (good/bad) reviews, we can the temper the results by assuming a beta(1,1)-distributed prior and then updating [1]. The expectation (which can thought of as an "adjusted average rating") ends up being:
E[x] = (good + 1)/(good + bad + 2)
This adjusts for situations where the number of reviews is small, in which a good average rating can be misleading.
[1] https://www.johndcook.com/blog/2011/09/27/bayesian-amazon/