4 ms·
Poisson, Pareto/power/zipf and normal distributions are really important. The top 3 for me. (What am I missing?) And often misused (most often normal). It’s r
by hammock 1y ago
Poisson, Pareto/power/zipf and normal distributions are really important. The top 3 for me. (What am I missing?) And often misused (most often normal). It’s really good to know which to use when
- klysm 1y agoNormal is overused for sometimes sensible reasons though. The CLT is really handy when you have to consider sums
- FilosofumRex 1y agoIt's surprising that so few people bother to use non-parametric probability distributions. With today's computational resources, there is no need for parametric closed form models (may be with the exception of Normal for historical reasons), each dataset contains its own distribution.
- klysm 1y agoIt’s easier to do MCMC when the distributions at hand have nice analytic properties so you can take derivatives etc. You should also have a very good understanding of the standards distributions and how they all relate to each other
- hyperbovine 1y agoHow hard is it to estimate that distribution for modern high dimensional data?
- jwarden 1y ago> What am I missing? Beta
- hammock 1y agoWhat are the common understandable use cases for beta distribution, in everyday life?
- jwarden 1y agoI don’t use probability distributions in everyday life ;) But it is the right distribution to represent uncertainty about the probability of binary events (eg a website user clicking some button). For example, if I have absolutely no idea the probability then I use the uniform distribution, Beta(1,1), which is the maximum entropy distribution. Then if I observe one user and they happen to click, I have Beta(2,1), and at a glance I known the mean of that (2/3) which is a useful point estimate.
- klysm 1y agoProportions of things frequently follow beta distributions. I think of it as the normal distribution of the domain 0 to 1.