A Bayesian Mixture Model with Application to Typhoon Rainfall Predictions in Taipei, Taiwan 1

In linear regression, it is typically assumed that the response variables are normally distributed. In practice, however, quite often the response variables are not only positive but with many zero measurements. In this article, we use Bayesian approach to analyze the data in which the distribution of the response variable is considered to be a mixture of a continuous distribution and a point mass at zero. Gibbs sampling algorithm is conducted to draw Bayesian inference and predictions and the results are quite accurate from the simulation study. The proposed model is also employed to make rainfall predictions for the typhoon data in Taipei, Taiwan.