| dc.contributor | 選舉研究中心 | |
| dc.creator (作者) | 俞振華;蔡佳泓 | zh_TW |
| dc.date (日期) | 2012 | |
| dc.date.accessioned | 19-Apr-2016 11:40:51 (UTC+8) | - |
| dc.date.available | 19-Apr-2016 11:40:51 (UTC+8) | - |
| dc.date.issued (上傳時間) | 19-Apr-2016 11:40:51 (UTC+8) | - |
| dc.identifier.uri (URI) | http://nccur.lib.nccu.edu.tw/handle/140.119/85563 | - |
| dc.description.abstract (摘要) | 過去數年來雖然網路普及率在台灣快速增加，各式各樣的網路調查在我們生活 周遭時常出現。但一般學術界對於網路調查的結果多半不會認真看待，因為網路調 查多半以非機率樣本的方式選擇受訪者，所得出的結果無助於我們推論母體特徵。 本研究計畫企圖延續去年研究者的提案，利用已建構完成的網路調查平台，試 圖探討利用網路調查方式時，因「非源於觀察此一動作所產生的誤差」所造成的估 計偏誤。主要的研究問題有以下兩個：首先，網路調查所造成的偏誤究竟有多嚴重？ 過去台灣並沒有系統性的探討究竟網路調查因樣本代表性不足所造成的偏誤為何。 本研究希望在這個問題上做出貢獻。利用電訪、面訪、與網路調查三方面的比對， 我們將可進一步判斷網路調查因樣本代表性不足所造成的偏誤程度。 而當我們確認因樣本代表性不足所造成的偏誤之後，我們應透過何種方式來校 正網路調查結果？本計畫將採事後加權處理的方式對於網路調查結果進行校正。其 中，仿效Lee（2006）及Lee and Valliant (2009)的方法，本研究將重點置於如何善用 「對照組」--即將網路調查樣本與面訪或電訪樣本相結合，產生新的樣本，再以入 選機率調整法（Propensity score adjustment）修正網路調查結果的偏誤。此外，本研 究亦將發展有如網路調查公司Harris Interactive 或Schonlau(2007)所提出的態度題 組，藉以區分網路使用者及非使用者，並利用該題組為橋樑，將平行的電話調查及 網路調查結合起來，以期日後可利用該方式減少傳統電話調查的樣本數，降低調查 成本。 | |
| dc.description.abstract (摘要) | While it is quite often to observe various web surveys in our day-to-day lives, scholars rarely consider their results seriously. It is mainly due to the fact that these survey results are products of op-in internet panels and can not be used to draw inference about the target population. This project aims to utilize the web survey platform built by the Election Study Center at National Chengchi University to study the possible estimation bias due to error of nonobservation in web surveys. The main subject of this project is twofold: first, it focuses on how serious the estimation bias could be in a given web survey. So far there exists no systematical analysis using Taiwan’s empirical web survey data to examine the extent to which nonrepresentative sample data can lead to estimation bias. Second, given the estimation bias of web survey, the next question would be how we can adjust it. In this project, we will pay special attention on using propensity score adjustment to overcome web survey’s noncoverage and nonresponse errors. Specifically, we will combine web survey samples with a “reference group” dataset (i.e., drawn from either face-to-face surveys or telephone surveys) to generate new samples and use propensity score adjustment (PSA) to produce proper estimation. Additionally, we will also develop the so-called “webographic” or attitudinal questions that tend to capture the differences between the online and the off-line population and can be used to bridge the data collected from telephone and web surveys. Again, by using the combined samples of phone and web respondents and “webographic” questions as predictors, we will apply PSA method to generate weights to adjust the estimation. | |
| dc.format.extent | 971619 bytes | - |
| dc.format.mimetype | application/pdf | - |
| dc.relation (關聯) | 計畫編號 NSC 101-2410-H004-079-SSS | |
| dc.subject (關鍵詞) | 網路調查;估計偏誤;樣本代表性;非源於觀察此一動作所產生的誤差;入選機率調整法 | |
| dc.subject (關鍵詞) | Web survey;Estimation bias;Sample representation;Error of nonobservation;Propensity score adjustment | |
| dc.title (題名) | 網路民意調查的建置與應用 | zh_TW |
| dc.title.alternative (其他題名) | Estimation for Web Surveys and Its Applications | |
| dc.type (資料類型) | report | |
