4 ms·
It's always worth checking the distribution. A 3 star rating with a normal-ish curve is very different than a 3 star rating with a lot of 5s and 1s -- and in t
by joelhoffman 8y ago
It's always worth checking the distribution. A 3 star rating with a normal-ish curve is very different than a 3 star rating with a lot of 5s and 1s -- and in that case the reviews will tell you clearly whose preconceived worldview the content fits.
- wenc 8y agoThat'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. [1] https://www.johndcook.com/blog/2011/09/27/bayesian-amazon/ https://www.johndcook.com/blog/2011/09/27/bayesian-amazon/