I was wondering how I could incorporate a prior to form a posterior distribution for multiple logistic regression.

More specifically, I am working with basketball data, where the response variable is binary (whether a player made the shot or not) and the explanatory variables are the distance from the basket and the defensive ratings of the opposing 5 players. I am able to generate multiple seasons worth of these shot logs for each player from the past, and wanted to know how I could take previous seasons of data into account when I also have a current season's worth of data.

  • $\begingroup$ In what sense is this* multivariate*? Do you have multiple response variables? $\endgroup$ Mar 15, 2016 at 8:41
  • $\begingroup$ Is that what multivariate means? Sorry I meant to say that there are multiple explanatory variables, but the response variable is either 0 or 1 $\endgroup$ Mar 16, 2016 at 17:19

1 Answer 1


This article discusses Bayesian logistic regression nicely. Basically, you have the flexibility to parametrize estimation however you see fit, but using a model which is linear on the log odds scale makes sense for many reasons. Furthermore, using a normal prior for log odds ratios should give you very approximately normal posteriors.

To do the sequential analysis, estimate the "prior" for this season by doing a first-pass Bayesian logistic regression for past seasons using a non-informative normal prior for that model.

  • $\begingroup$ Thanks for the response Adam. But what if my covariates are continuous and not discrete? $\endgroup$ Mar 19, 2016 at 22:00
  • $\begingroup$ @kleiza don't put a prior on your covariates. We generally think of them as "fixed" or given in regression modeling. $\endgroup$
    – AdamO
    Mar 19, 2016 at 22:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.