Hi
I have hierarchical model
first level
data are y_i~ Normal (alpha_i, S_i2 )
where y_i , S_i2 are known.
second level
alpha_i~Normal(alpha, tau2)
third level
priors of alpha and tau are
alpha~ Normal(0,10000)
tau ~ halfNormal (0,100), tau > 0
I used Metropolis within Gibbs sampler
I use normal distribution as proposal distribution cantered around current value of tau and different values of standard deviation and I choose candidate value that is greater than zero.
The problem with acceptance rate is very low and when I choose smaller values of proposal standard deviation I get good acceptance rate but the mixing of chain is very bad.
I have attached pdf file that has more detail and contains my code.
I do not know if my code is not right or there is another thing wrong?
Thank you
- Badges
- Science topic
- Similar topics
- Social Psychology
- Values