查詢結果分析
相關文獻
- 邏輯斯迴歸'擬'最大概度估計值之漸近性質
- 在複雜抽樣下邏輯斯迴歸模式之適合度檢定
- 臺灣地區零至六歲兒童體位標準調查抽樣設計之研究
- 駕駛行為與肇事概率之關聯性研究
- Logistic Regression Analysis of Data from a Cluster Sample
- 邏輯斯迴歸模式與對數線性模型之比較
- Exploring Willingness and Related Factors by Using the Healthcare Report Cards for hospitals in Taiwan
- 分層母體中之調適型集群抽樣
- 法拍屋貸款授信違約研究--考量投資客特性之判斷模型
- 大專籃球選手情緒智力、壓力因應策略對運動成績影響之研究
頁籤選單縮合
題名 | 邏輯斯迴歸'擬'最大概度估計值之漸近性質=The Asymptotic Properties of 'Pseudo' Maximum Likelihood Estimates for the Logistic Regression Model |
---|---|
作者姓名(中文) | 沈葆聖; | 書刊名 | 中國統計學報 |
卷期 | 35:3 1997.09[民86.09] |
頁次 | 頁271-277 |
分類號 | 319.51 |
關鍵詞 | 集群抽樣; 邏輯斯迴歸; 設計效應; Cluster sampling; Logistic regression; Design effect; |
語文 | 中文(Chinese) |
中文摘要 | Rao and Scott (1992)從抽樣設計效應(design effect)和有效樣本大小( effective sample size)的觀點提出'擬'最大概度估計值('pseudo' maximum likelihood estimate) □,用以估計集群抽樣(cluster sampling)下邏輯斯迴歸(logistic regression)模式之參數。本文推導□的漸近性質,並證明□比二項式分佈假設下之最大 概度估計值□漸近有效。從證明中並得知□相對於□的效率隨著子族群(subpopulation) 設計效應之變異數□的增加而加大。模擬結果顯示在有限樣本下,當□小時,□相對於□ 的優越性不再成立。 |
英文摘要 | Based on the concepts of the design effect and effective sample size in sample survey, Rao and Scott (1992) proposed 'pseudo' maximum likelihood estimate, □, for estimating the logistic regression parameters under cluster sampling. In this article, the asymptotic properties of □ are derived and □ is shown to be asymptotically more eficient than maximum likelihood estimate, □, under the assumption of independent binomial sampling. From the proof, it can be seen that, relative to □,□ becomes increasingly more efficient as the variance of the design effect among subpopulations, □, increases. Simulation results demonstrate that, for finite samples, the relative superiority of □ over □ does not hold when □ is small. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。