以廣義估計方程式分析影響陽明山地區之降雨量因素

Abstract

當個體只有一個觀測值，廣義線性模型(McCullagh& Nelder, 1983)已經被廣泛的用來求得參數估計；當個體不只一個觀測值，觀測值趨近於常態分配，這種統計模型已經提出(Laird & Ware, 1982)；當我們是離散型觀測值，統計方法也被多人提出(Ochi & Prentice, 1984. Laird & Ware, 1984. Koch, 1977)；當個體不只一個觀測值，並且觀測值非常態時，這種統計方法是廣義估計方程式(Liang & Scott L, 1986)。此篇論文分析竹子湖觀測站的氣象觀測資料，從廣義線性模型進一步延伸，利用廣義估計方程式分析長期追蹤資料，再利用廣義估計方程式中模型選擇的方法QIC，選取適合的模型，以實際例子進行操作說明。

If each subject has only one observation, then the generalized linear model (McCullagh& Nelder, 1983) had been used to obtain parameter estimate. If each subject has more than one observation and observations tend to follow normal distribution, this statistical model has been proposed (Laird & Ware, 1982). If we have discrete observations, then this statistical model also has been proposed (Ochi & Prentice, 1984. Laird & Ware, 1984. Koch, 1977.). If there is more than one observation and observations do not tend to follow normal distribution, this statistical model is GeneralizedEstimating Equations(Liang& Scott L, 1986). This thesis analyzes the meteorological observation data of Zhuzihu and we use the generalized estimation equation to analyze the longitudinal data. And then we use the method of model selection of generalized estimation equation, QIC, to choose the best model. We take an example by using this method of model selection.