蔡蓉青呂翠珊Tsai, Rung-ChingLu, Tsui-Shan林克謙Lin, Ke-Chian2019-09-052015-3-42019-09-052015http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22GN060040032S%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/101699一份問卷調查的健全推論建立在有效率的估計之上，但是有效率的估計往往受到不認真的受訪者的影響。問卷調查廣泛應用虛題(bogus item)找出不認真的受訪者、發現那些不看題目的人。另一種方法是反應答題的時間，其使用方式已由Rasch混合模型結合反應時間的方法確立，並能有效地提升估計的效率。在這篇研究中，我們提出結合虛題與混合反應時間的模型。此模型是利用虛題並考慮反應時間來區分「兩種不同類型的不認真受訪者的回答以及認真受訪者的回覆」。分析時保留虛題做為共變數與時間做為潛在類別的應變數的優點。在模擬實驗中，當樣本數500人以上，反應時間有降低估計值的標準誤的趨勢。更進一步，我們應用該模型分析真實資料時，其中有高達22\%的不認真受訪者。結果顯示：我們未受限制的模型表現得比受限制的模型好。A valid inference for a survey is always based on efficient estimation. The efficient estimation is threatened and distorted by the careless response. Bogus item is commonly used in survey questionnaire to detect the careless respondents and powerfully unveils those responding without reading the items. Response time is another way to identify the careless responses and its use of enhancing item parameter estimation was verified by the mixture Rasch model with response time component (MRM-RT). Therefore, we propose a mixture response time model to analyze survey questionnaire with a bogus item. The goal of our proposed model is to classify two behaviors of the careless responding in addition to the attentive respondents, using the bogus item and taking account into response time. We attempt to converse the benefit of the bogus item used as a covariate and the response time used as the latent class indicators in analysis. In the simulation studies, we find that the standard errors of the model with response time decrease substantially as the sample size is equal or larger than 500. Further, we also apply our proposed model to a real data set and with a larger proportion of careless responding (22\%), the results show that the unrestricted models perform better than the restricted models.虛題反應時間潛在類別Rasch混合模型結合反應時間bogus itemresponse timelatent classMRM-RT