From the results obtained for the artificial GARCH data we found that the importance sampling outperforms the
MCMC method implemented by the MH algorithm. We also performed the Bayesian inference for the daily stock
return data of Toshiba Co. and found that the performance of importance sampling is comparable to that of the
MCMC method. Our results might indicate that the performance of two methods depends on the return data we
consider. Since there exists a case that the performance of importance sampling is better than that of the MCMC
method, it is concluded that the importance sampling method can be used as an alternative method to the MCMC
method for the Bayesian inference of the GARCH model.