Using M-H algorithm I want to generate samples from proposal distribution which will ultimately converge to the target distribution.
I am facing a problem particularly in updating the prior to posterior and using posterior as prior again to find new posterior.
I am not using a conjugate prior. So while updating:
post1=likelihood1*prior;
post2=likelihood2*post1; =likelihood2*likelihood1*prior;
post3=likelihood3*likelihood2*likelihood1*prior;
this keeps on going which seems not feasible if I am generating N number of samples.
Hiii,i am also working on bayesian mcmc and got stuck with the same problem.if u get answer will u please help me out.
Hi!
I hope you are sequentially getting more data and use this new data while calculating likelihood_{i}?
I have got the solution to this problem!!!!
You dont have to sequentially update the posterior. You just need to sample from Likelihood times prior which will give you a target distribution and later use MCMC-MH algorithm to sample from the target
I hope I am clear. If any one has any doubt do post your queries in this forum.
Hi Nikil
I think I have the same problem. Is your problem as follow?
you have e.g. three given data set, then you wanted to construct three likelihood functions? next, you created the posterior as post3=likelihood3*likelihood2*likelihood1*prior;
if it is so, are the system responses you used to create the likelihood, independent?
- Badges
- Science topic