2005-10-012024-05-17https://scholars.lib.ntu.edu.tw/handle/123456789/688651摘要：本研究主題是有關因時間改變而喜好程度轉變之數學模式應用。此研究之架構是屬於一種以排比關係為基礎之媒介理論。媒介之概念抓住系統內諸排比（稱之為‘狀態’）之間經由信號傳遞的轉換關係。本研究將討論一種以semiorder為基礎之隨機遊走模式。此模式是媒介理論下之一特例。此模式之可行處在於其對於諸(隱含)狀態之&#23515;鬆及符合人性之假設。在此之前類似模式曾被發展及應用於1992年美國總統選舉之資料分析上，但是此應用並不成功。我們計劃用不同的方式來發展此以semiorder為基礎之隨機遊走模式，而且將應用於同一筆資料。 由於美國總統選舉之資料是以量數評定為依據，而且此評定與以semiorder為基礎之排比關係並無一對一之轉換關係，本計劃的重點之一便在於此轉換關係之建立。特別的，本計劃將發展並詳細列舉一套以機率為主之反應機制來敘述此量數評定與以semiorder為基礎之排比之間的關係。本計劃並將持續原有研究，探討以semiorder為基礎之隨機遊走模式的延伸母模式—所謂的轉入轉出模式。這些模式並將應用於1996 年及2000年美國總統選舉之資料分析上以驗證其有效性。本計劃之部分工作將致力於不同子模式間之建構，分析，比較，及以適合度為基礎之統計假設考驗。此計劃之結果預期將有助於理解選民對候選人喜好程度改變的誘因。另外值得一提的是本計劃所採用的研究方法亦對國內選舉研究提供了一個新的思考方向。<br> Abstract: This research topic is motivated by the modeling of the evolution of preferences over time. The general framework of such research is the `media theory.` The concept of a `medium` captures the organization of a collection of `states` through a set of transformations, or `tokens,` each of which maps the set of `states` into itself. A special case of the media theory is the random walk model based on semiorders. The interest of such model is that it makes less constraining assumptions concerning the (latent mental) states. However, previous attempt to apply the semiorder model to the 1992 U.S. presidential election panel data has failed. Accordingly, the present project intends to use a different approach to apply the semiorder model to the same data set. Knowing that the panel data consist of thermometer scores to which there are no one-to-one correspondences to their counterparts of semiorder rankings, we aim to investigate various probabilistic models in which the `states` of the medium are semiorders but the respondents are required to respond using thermometer scores. Especially, the present project aims to spell out the response mechanism(s) specifying how a respondent will translate a semiorder preference into numerical thermometer scores. The specified semiorder model and its tune in-and-out extension will then be applied to the 1992 U.S. presidential election panel data. For the purpose of cross validation, the present project also intends to apply the semiorder model, as well as its tune in-and-out extension, to the 1996 and 2000 data sets. A substantial amount of work will be devoted to the comparison of multiple models and the study of a range of different substantively motivated hypothesis tests on multiple data sets. The results will shed light on the understanding of the possible sources of (tokens of) information that might affect the voter’s preferences. Especially noteworthy is the research methodology used in the present project: it might provide a different angle of thinking toward the election studies in Taiwan.態度改變總統選舉隨機遊走排比關係數學模式媒介理論Attitude changeMedia theoryPanel dataPersuasionPresidential electionRandom walkSemiorder