Please use this identifier to cite or link to this item: https://ah.lib.nccu.edu.tw/handle/140.119/146226
題名: 雙底冊電話調查合併組合加權處理之比較分析
A Comparative Analysis of Weighting Approaches for Dual-Frame Telephone Surveys
作者: 侯佩君
Hou, Pei-chun
貢獻者: 選舉研究
關鍵詞: 雙底冊電話調查; 加權; 設計效果; 加權損耗; 偏差
dual-frame telephone surveys; weighting; design effect; weighting loss; bias
日期: May-2023
上傳時間: 13-Jul-2023
摘要: 隨著科技進步與時代變遷,近年來臺灣面臨單底冊電話樣本的涵蓋誤差日益增大問題,不論市話調查或手機調查;而雙底冊電話調查的施行應可有效降低此一缺失。惟有關雙底冊電話調查的資料合併策略與加權處理,當下臺灣學界仍莫衷一是。本文旨在利用臺灣學界甚少觸及的變異數與偏差,以2020年雙底冊電話調查的實務資料,來評比各類型雙底冊調查的合併組合在各種加權策略下的加權效果,包括重疊設計與篩選設計;此外,本文也進一步探討非等比例樣本配置的雙底冊電話調查在資料推估上的差異,提供實務應用參考。 本文分析結果顯示:1.在雙底冊各自具有完整代表性且樣本數近乎等比例的條件下,採用一般的母體人口特徵變項來校正加權,不論是使用重疊設計或篩選設計來合併,因加權而增加的變異並不大,且僅事後分層組合估計的樣本估計值與其他合併組合有所差距。2.若雙底冊的樣本採用不等比例配置,以一個完整代表性底冊搭配另一個僅具有1/3或1/2樣本數的底冊,利用篩選設計、重疊設計的直接合併及有效樣本數合併等三種方式進行比較後發現,因應加權所衍生的加權耗損不大,而彼此之間在樣本估計值上的差異也很小。
With the rapid advancement of IT and the changing times, Taiwan has been facing increasing coverage error in single-frame telephone surveys (whether by landline or cell phone) in recent years. Dual-frame telephone surveys can effectively reduce this deficiency. However, there is still no consensus on how to weight dualframe telephone samples in Taiwan. This study uses data from a 2020 dual-frame telephone survey to compare several popular weighting approaches. The impact of these approaches on the variance and bias in both single-frame and dual-frame samples, as well as in overlap and screener designs, is discussed. Additionally, this study examines the differences between the estimates of the non-proportion dual-frame sample allocation to provide practical recommendations. The results indicate the following: (1) if the sample allocation of the dual-frame samples is in nearly equal proportions, minor differences occur in the weighting loss between the overlap and screener designs. Only the post-stratified estimating procedure reveals some differences in the estimates with other designs. (2) If the sample allocation of the dual-frame samples is in unequal proportions, such as one single-frame national sample and the others are only 1/3 or 1/2 of the former, then after calibration, the weighting loss is only slightly increased, and these approaches produce only tiny differences in the calibrated estimates between the combinations.
關聯: 選舉研究, 30(1), 43-88
資料類型: article
DOI: https://doi.org/10.6612/tjes.202305_30(1).0002
Appears in Collections:期刊論文

Files in This Item:
File Description SizeFormat
30-1-2.pdf2.04 MBAdobe PDF2View/Open
Show full item record

Items in NCCU Academic Hub are protected by copyright, with all rights reserved, unless otherwise indicated.