I've been running a glm with a binomial distribution and two predictors which are date and a categorical variable named 'pond':

glm(yes/no ~ date + pond, family=binomial(link="logit"), data=dat)

However, running into trouble as 'pond has 15 levels (leading to quasi separation). Would running a bayglm be appropriate to deal with this issue (I cant reduce the number of levels)? If so, how would I select appropriate priors, given that I don't have any prior/expert data. Are there certain 'default' priors I should use, considering it's a binomial glm?

  • $\begingroup$ I am voting to leave open because the questions seem to be "Is the method I mention here appropriate?" and "How should I choose a prior in the absence of data/expert judgement?". These sound like conceptual questions, not programming ones. $\endgroup$ – mkt Jan 29 at 10:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.