Hacker News new | past | comments | ask | show | jobs | submit login

Yes, single scalar values tell you nothing useful if you have no a priori knowledge of the shape of your distribution. Mean and variance are meaningless figures if your distribution is multimodal, for example. If you truly have to compress the distribution into a few numbers, then the best thing to do is to represent it as a series of quantile values. In cases where the distribution is unknown, evenly spaced quantile values are a good start.





Consider applying for YC's Summer 2025 batch! Applications are open till May 13

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: