測量誤差模式中的延伸校正分數函數

淡江大學機構典藏 > 理學院 > 應用數學與數據科學學系 > 研究報告 > Item 987654321/101289

請使用永久網址來引用或連結此文件: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/101289

題名:	測量誤差模式中的延伸校正分數函數
其他題名:	Extensively Corrected Scores in Measurement Error Models
作者:	黃逸輝
貢獻者:	淡江大學數學學系
關鍵詞:	測量誤差模式;校正分數函數;條件分數函數;延伸校正分數函數;measurement errors;corrected score;conditional score;extensively corrected score
日期:	2013-08
上傳時間:	2015-04-21 14:35:08 (UTC+8)
摘要:	在測量誤差模型的分析方法中，傳統上只有校正分數函數及條件分數函數的結果具有一致性，然而它們的應用受到許多限制；例如校正分數函數的存在要求被估計函數必須是一個複數平面上的entire function，而條件分數函數則需要能找到缺失自變數的充分統計量，這些假設限制了這兩個方法的應用範圍。本計畫預計發展一種新的估計方法，我們並不打算直接估計原始的分數函數，因為該函數可能無法被不偏地估計 (Stefanski,1989)，反而是估計加權後的分數函數，得到”加權的分數函數”的估計量後我們再利用權重的近似值將結果”反加權”回來，最後所得的函數我們稱為延伸校正分數函數。我們預期延伸校正分數函數除了有一致性外也會有不錯的效率，它也可以應用於條件分數函數及校正分數函數皆不適宜的情況。 There are two major consistent estimations in measurement error problems. The conditional score and the corrected score. The former method requires a sufficient statistic for the true covariate can be found, while the later method requires that the object score function can be estimated unbiasedly by surrogates and responses. Due to these assumptions, they are not applicable even for some simple models. For example, they are not applicable for the logistic regression with quadratic predictor. In this project, we will develop a novel method for estimating the object function. In stead of directly estimating the original score function which can be difficult or impossible, we will estimate a weighted version of it, and then reweigh it back for efficiency concern. Though the resultant function, the extensively corrected score, is not an unbiased estimator of the original score function, it has 0-mean property and converge to the original score function when measurement error approaches 0. It should be more flexible and applicable than the two traditional methods. We expect it to yield consistent estimates with good efficiencies.
顯示於類別:	[應用數學與數據科學學系] 研究報告

文件中的檔案:

沒有與此文件相關的檔案.

在機構典藏中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....