學(xué)術(shù)沙龍主題: The Poisson Item Count Technique and its non-compliance design for survey with sensitive questions.
報告人: 吳琴，華南師范大學(xué)副教授
報告時間: 2025年4月2日(周三)；上午9:00—10:30
報告地點(diǎn)：南校區(qū)網(wǎng)絡(luò)安全大樓 120 會議室
報告人簡介：吳琴，博士，畢業(yè)于香港浸會大學(xué)統(tǒng)計(jì)系，現(xiàn)于華南師范大學(xué)統(tǒng)計(jì)系工作, 副教授。現(xiàn)主持國家自然科學(xué)基金面上項(xiàng)目1項(xiàng)（在研），青年項(xiàng)目1項(xiàng)(已結(jié)題)，廣東省質(zhì)量工程項(xiàng)目1項(xiàng)(已結(jié)題)。相關(guān)研究成果被統(tǒng)計(jì)雜志Statistical Methods in Medical Research， Statistics in Medicine，Biometrical Journal，Statistical Papers等雜志收錄。
報告摘要: The Poisson item count technique (PICT) is a survey method that was recently developed to elicit respondents’ truthful answers to sensitive questions. It simplifies the well-known item count technique (ICT) by replacing a list of independent innocuous questions in known proportions with a single innocuous counting question. However, ICT and PICT both rely on the strong “no design effect assumption” (ie, respondents give the same answers to the innocuous items regardless of the absence or presence of the sensitive item in the list) and “no liar” (ie, all respondents give truthful answers) assumptions. To address the problem of self-protective behavior and provide more reliable analyses, we introduced a noncompliance parameter into the existing PICT. Based on the survey design of PICT, we considered more practical model assumptions and developed the corresponding statistical inferences. Simulation studies were conducted to evaluate the performance of our method. Finally, a real example of automobile insurance fraud was used to demonstrate our method.
