頁籤選單縮合
題名 | 水文頻率分析之離群值、可信區間與最小需求記錄年數=Outliers, Confidence Intervals and Minimum Required Length of Records in Hydrological Frequency Analysis |
---|---|
作者姓名(中文) | 王如意; 李孟穎; | 書刊名 | 臺灣水利 |
卷期 | 47:3=187 1999.09[民88.09] |
頁次 | 頁1-21 |
分類號 | 443.046 |
關鍵詞 | 離群值; 最小需求記錄年數; 線性動差法; 褶刀法; 靴環法; Outlier; Minimum required length of records; Linear moments; Method of jackknife; Method of bootstrap; |
語文 | 中文(Chinese) |
中文摘要 | 本文嘗試對水文頻率分析中有關量值推估、離群值處理、可信區間及最小需求記 錄年數推求時所產生之若干疑點加以探討。 文中比較乘積動差法與線性動差法之特性,選擇其一作為機率分布函數之推估方法,再結合 褶刀法與不協和測度作離群值之選取,次以靴環法為基礎建構虛無元素之觀念作離群值之判 定,又以靴環法作可信區間之推估,並藉由可信區間之推估,求得特定推估量在一定準確範 圍內之最小需求記錄年數。 本文除使用理論模擬值作頻率分析方法之驗證外,亦採用五堵站、阿里山站及臺南站之雨量 資料作為此分析方法之應用實例,藉以顯示五堵站及阿里山站水文紀錄之離群值存在性,並 建議該離群值之取捨方式,以求取特定推估量之可信區間,其為滿足某準確範圍條件下,記 錄資料應具有之最小需求記錄年數。 |
英文摘要 | This study is carried out for better understanding of some controversial topics about estimation of quantiles, dealing with outliers and evaluations of both confidence interval and minimum required length of records of hydrological data. Parametric estimation method is chosen by comparing between two categories: the method of product moments and the method of linear moments. Based on this parametric estimation method, the most probable outlier can be preliminarily selected using the method of jackknife combined with a measure of discordancy, and then tested for its outlying by utilizing the null element proposed in this study based on the method of bootstrap. After that, the confidence interval can be constructed and the minimum required length of records may be evaluated for the estimated quantile under the certain tolerance limits. Theoretical simulations are given for verification of the method proposed. Results of analysis on rainfall data from Wutu, Alisan and Tainan stations are then taken as examples, by which the existence of outliers in data from Wutu and Alisan stations can be shown, and their acceptance or rejection proposed. The minimum required length of records under some tolerant precision is decided then, along with the confidence intervals for the estimated quantity. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。