I am working on parameter estimation of GARCH model with Metropolis Hastings. But the results I have got doesn't look reasonable, actually it is quite different from what I have got from Gibbs sampler. Since there is some upper bound constraints on some of the parameters, I use truncated Normal distribution(so that sample won't be drawn from outside the constraints) as the proposal distribution. Will this be the reason that the MH is not working well? But if I use a Normal with constraint, quite a number of samples will be discarded right away, plus it is not converging after 10k draws. I understand that the proposal distribution will impact the acceptance rate, but don't understand how distort the estimation. Any comment will be helpful.
Solved – How to incorporate parameter constraints in Metropolis Hastings
garchgibbsmarkov-chain-montecarlometropolis-hastingssimulation
Best Answer
If you use a truncated normal as your proposal, $\mathcal{N}_a^b(\mu,\sigma^2)$ where $(a,b)$ is the truncation interval,