Created
December 29, 2015 12:31
-
-
Save ruescasd/5f11220088f45e80d422 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Modeling a binomial process with the beta binomial | |
Case 1: You have no information | |
This requires using a non-informative prior for the beta distribution. Normally this is done | |
with a uniform prior for all values of p, which corresponds to alpha = beta = 1. | |
Case 2: You have some prior information in terms of the mean and variance of p | |
You calculate alpha and beta, like in this post: http://stats.stackexchange.com/questions/12232/calculating-the-parameters-of-a-beta-distribution-using-the-mean-and-variance | |
That gives you a beta prior which reflects your prior information. | |
Case 3: You have information from previous binomial events, in the form ok k successes out of n trials: | |
Given that the beta is conjugate to the binomial, the posterior takes the form of the previous beta like this: | |
Beta(alpha, beta) => Beta(alpha + k, beta + n - k) | |
Thus, if you started with a uniform prior and have info on binomial events, you get this prior: | |
Beta(alpha, beta) => Beta(1 + k, 1 + n - k) | |
The process is, prior + evidence = posterior, which becomes the new prior for your new distribution. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment