演講公告
新聞標題: ( 2023-12-01 )
演講主題:Stable estimation of time-varying treatment effects using an approximate multiply robust estimator
主講人:田栗正隆教授 (日本東京醫科大學)
演講日期:2023年12月19日 (二) 11:00 - 12:00
演講地點:光復校區 綜合一館四樓427室
摘要內容:
報名表單 (有意參加者請協助填寫報名表單,以利場地安排,謝謝您的配合)
https://forms.gle/nzwnU6Hcn7H9yo8u7
Abstract
In longitudinal studies involving time-varying treatments, time-dependent confounding can occur if the time-dependent covariates are influenced by past treatments. Multiply robust estimators, which are based on augmented inverse probability (AIPW) estimating equations, have been proposed to protect against model misspecification(Bang and Robins, 2005; Rotnitzky et al., 2017). However, these estimators have large variances when there is significant variability in inverse probability weights. To address this issue, we propose an approximate multiply robust estimator based on stratification of inverse probability weights.For the point treatment, our proposed method corresponds to the approach combining outcome regression models and propensity score stratification (Lunceford and Davidian, 2004). We will present the results of simulations comparing the proposed estimator with existing estimators.相關檔案:Postor.jpg