I read that uninformative priors for population variances are often represented by invgamma(eps,eps) where eps could be 1, 0.1 or 0.001.

In my model I used these but variance sometimes goes upto infinity. Due to this, the prior of mean which is a truncated-normal distribution is disturbed as sample values from this prior variance distribution exceeds the expected variance. Thus, my MCMC-metropolis hastings-gibbs sampler failed to converge or there were some errors.

So, now I have used invgamma(2,2) as uninformative prior for variance because the sample values are not too high. But, is this distribution allowed for describing "uninformativeness"?

Please help, I am having a tough time making my bayesian inference - MCMC model converge to correct posterior distribution.


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.