造成颱風的原因很多，影響其降雨的因素也很複雜，導致颱風資料充滿隨機性，統計學識解釋隨機性的一種工具。本計劃欲利用統計方法就中央氣象局已收集近數十年的資料進行颱風雨量之資料分析，我們將繼已建立之適當迴歸模型，考慮以氣象專家之相關意見建立先驗分布，利用貝氏分析進行降雨量之預測模型。同時亦以類似方法發展風速預測之適當模型。考慮的模型除一般線性迴歸外，尚探討分析類別性資料之羅吉斯迴歸模型，以利氣象局日後針對雨量分級制度和風級數之預報。本年度將首次直接針對資料之相關性以長時期資料分析方法建模，並進行分析。 Due to the complexity of typhoon structure, the resulting rainfall data are completely random. Statistics is a tool to explain random phenomena. We propose to apply statistical methods to analyze the typhoon rainfall data collected by the Central Weather Bureau over the recent decades. The major concern will be on statistical data analysis as well as efficient statistical forecasts. We will continue to study the Bayesian analysis based on the regression model developed in previous project by incorporating the related information provided by the meteorologists as the prior information into the model for rainfall and windspeed predictions. In addition to the usual linear regression model, we will furthermore employ the logistic regression model which is used to analyze the categorical data for rainfall and wind speed prediction. Hopefully, it can help the CWB to enhance the prediction ability on rainfall and windspeed levels. In addition, we will consider directly modeling and analyzing the data based on their dependent structure via the longitudinal data models to improve the forecast. 研究期間：9401 ~ 9412
