二氧化硫一直是台灣地區空氣污染中主要污染物之一，本研究將二氧化硫的時空資料，表示為數個基底函數與平穩過程的線性組合，可把平均數與變異數估計問題視為迴歸分析，再利用 Tibshirani (1996)所提出的「最小絕對壓縮挑選機制」(Least absolute shrinkage and selection operator，Lasso)，來選取基底函數及估計參數，分析結果以圖的方式呈現二氧化硫之空間分佈。總體而言，西半部地區在春季(五月)時，空氣中二氧化硫含量會上升偏高，而整個東半部及離島則全年皆低；至於各地區標準差在桃竹、雲林及高屏等三大工業區皆偏高。 SO2 has been one of the major air pollutions in Taiwan area. In this research, the spatial-temporal data of SO2 is decomposed as a linear combination of basis functions and stationary processes. The problem of mean and variance estimations can be considered as a regression. A subset selection method, least absolute shrinkage and selection operator (Lasso), proposed by Tibshirani (1996) is used to choose a suitable subset of the basis functions and estimate the parameters. The analysis results are demonstrated in graphs which can be easily observed the spatial distribution of SO2 and the seasonal variation of it. The monthly mean of SO2 is increasing in May over western Taiwan. As in eastern Taiwan and off-shore islands, the monthly mean of SO2 is very low through the year. High values of the standard deviation in SO2 occur in the surrounding area of three major industrial zones.
智慧科技與應用統計學報=Journal of Taiwan Intelligent Technologies and Applied Statistics 11(1)，頁25-34