A browser extension which adds confidence intervals to Amazon ratings.
- Inspired by this 3blue1brown video: https://youtu.be/8idr1WZ1A7Q
- If the ratings were binary, we could just use the Beta distribution to estimate the lower and upper bounds of a 95% confidence interval, but Amazon ratings are out of 5 stars.
- Maybe we can generalize the problem from binomial to multinomial, and use a Dirichlet distribution instead to get our confidence interval (see: https://stats.stackexchange.com/questions/15979/how-to-find-confidence-intervals-for-ratings)? But would that assume independence of each of the star ratings? Is that a fair assumption?
- Maybe bootstrapping would be easier, but it's computationally expensive when there are a very large number of ratings and ineffective when there are a very small number of ratings.
- This also doesn't take into account fake ratings; we should ignore that concern for now.
- Amazon groups seller ratings into positive (4 or 5 starts), neutral (3 stars), and negative (1 or 2 stars) ratings (see: https://www.amazon.com/sp?seller=A3EPIN8KW1GGQX), so as a workaround, we can treat "probability of positive experience" as probability of having a 4 or 5 star rating, and just get a binomial confidence interval on that.
First, make sure you have Node.js and either NPM or Yarn. Then install the dependencies.
npm install
yarn installRight now, we are working on the Chrome extension. We welcome pull requests for other platforms!
To build the Chrome extension, simply run the build command. The extension will appear in dist/chrome.
npm run build:chrome
yarn run build:chromeIf you would like to automatically rebuild the extension each time files are modified, run the watch command.
npm run watch:chrome
yarn run watch:chromeTo build a production-ready extension (for end users), add --env.production to the build command.
npm run build:chrome --env.production
yarn run build:chrome --env.production