Survey Research - Method and Application
logo-Survey Research - Method and Application

Survey Research - Method and Application
logo_m-Survey Research - Method and Application

    Jump To中央區塊/Main Content :::
  • About
  • For Authors
    • Submission Guidelines
    • Review Process
    • Editorial Policy
    • Online Submission
  • Editorial Board
  • Issues
  • Ethics
  • Contact
  • RCHSS
中文
RCHSS
search
Survey Research-Method and Application
TSSCI Tier 1
  • Home
  • Issues
  • April, 2005 "Survey Research—Method and Application" Volumn 17
  • Facebook
  • line
  • email
  • Twitter
  • Print
2005 / April No.17
Using Samplemiser to Predict Presidential Elections: Studies of Taiwan and the United States
發刊日期/Published Date
2005 / April
中英文篇名/Title
時序模型在選舉預測上的應用:以美國、台灣總統選舉為例
Using Samplemiser to Predict Presidential Elections: Studies of Taiwan and the United States
論文屬性/Type
研究論文 Article
作者/Author
徐永明
Yung-Ming Hsu
頁碼/Pagination
111-149
摘要/Abstract

民意調查目的在於使用電訪所獲得的資訊進行預測,然而民調資訊有新有舊,那麼在預測時到底應該用哪一次民調的結果進行預測呢?因此本文將以Samplemiser(Green et al.1999)為例,希望能將多次民調資訊皆運用在選舉預測上,並且在時序模型的分析下,以Kalman filtering and smoothing algorithm(卡爾曼濾子與平滑估計值)將民意調查中真實的民意變動從抽樣誤差所產生的變動中區分出來,更進一步運用卡爾曼濾子與平滑估計值可將各個民調資訊進行跨越時間結合分析之性質,累積各個民調下的資訊,以提升民意調查預測的精確性。最後我們以台灣與美國2000年以及2004年總統大選的資料進行實證分析,發現當樣本數目越小或是使用的民調次數越少時,卡爾曼濾子與平滑估計值將可以越有效的降低抽樣誤差,而我們也可以獲得一個較無偏差且平穩的指標進行預測。以2004年陳水扁在台北市的支持度為例,原本約落在10%至40%的區間內,其離散程度頗大,然而在卡爾曼濾子與平滑估計值的估計下,我們可以獲得一個較無偏差且平穩的指標,以預測陳水扁在台北市市民的支持度。這樣的結論可以使我們將多筆資料進行時序性的分析,其對於選舉預測的幫助將遠勝於單筆民調資訊的選舉預測能力。

This paper aims to examine the change and stability of public support amongst electoral campaigns. Using the data collected by various polling before election day, I sort out the signal from estimated noise to specify the pattern of mass alignment in presidential election. Applying the Green's (Green et al.1999) Samplemiser, I uncovered the reduction of estimated error across the polling data and contended that using Samplemiser is an appropriate way to re-estimate the mass attitude, and to predict the electoral result before election day.

關鍵字/Keyword
民意調查, 卡爾曼濾子, 總統大選,
Public Opinion, Kalman Filtering, Presidential Election, Samplemiser
學科分類/Subject
政治學
Political Science
主題分類/Theme

DOI
https://doi.org/10.7014/SRMA.2005040003
檔案下載/Download
Full-text
  • About
  • For Authors
  • Editorial Board
  • Issues
  • Ethics
  • Contact

128 Academia Road, Section 2, Nankang, Taipei 11529, Taiwan

Tel: +886-2-2787-1816 Fax: (02)2788-1740 Email: srcsr@gate.sinica.edu.tw

© Copyright 2026. RCHSS Sinica All Rights Reserved.Privacy Policy & Security PolicyVersion:V1.1.3