This should be straightforward. I come from a frequentist background and I'm trying to understand bayesian logic. If I use a flat prior for my parameter to allow "the data to speak", what's exactly the benefit of bayesian over frequentist? As I understand it, I'm not allowing any shrinkage so I'm not really using any prior judgement. Obviously I'm missing something, can anyone help?
Asked
Active
Viewed 421 times
2
-
No matter what prior you use the Bayesian interpretation involves treating the parameter as a random variable with a posterior distribution. Jefferies studied flat priors. Some are improper which means they are not actual distributions. Most Bayesians advocate informative priors and in the old days they used conjugate priors for ease of computations. The situation is very involved and you should do more reading. – Michael R. Chernick Nov 26 '17 at 23:52
-
There are many similar questions on X validated about [non-informative priors](https://stats.stackexchange.com/a/20535/7224), questions that you should try to relate to so that you can refine your question into something less vague. – Xi'an Nov 27 '17 at 04:47