Bayesian extreme for modeling high PM10 concentration in Johor

The aim of this study is to determine the behavior of extreme PM10 levels monitored at three air monitoring stations in Johor using frequentist and Bayesian technique. Bayesian allows priors or additional information about the data into the analysis which expectedly improve the model fit. The generalized extreme value distribution is fitted to the monthly maxima PM10 data. The results obtained show that the Bayesian posterior inferences perform at least as trustworthy as maximum likelihood estimates but considerably more flexible and informative. The return levels for 10, 50 and 100-years were computed for future prediction