This is a well-reasoned exercise about constructing confidence intervals for the population rating for a product from individual ratings observed thus far. It doesn't do well with bi-modal (e.g. 100 `1` ratings and 100 `5` ratings) but then I don't think most methods do. Arguably, the confidence interval in that case ought to cover the entire range of values.
Apart from that, this is one of that rare breed of Stats related posts that correctly describes the concept of a confidence interval/set. I do not know why it was flagged. It should not have been.
Apart from that, this is one of that rare breed of Stats related posts that correctly describes the concept of a confidence interval/set. I do not know why it was flagged. It should not have been.