Time-varying treatmentsに対するIPTW法による 因果効果の推定 ○鈴木 徳太1、折原隼一郎2 1横浜市立大学 データサイエンス学部 2東京医科大学 医療データサイエンス分野 Estimation of causal effect on time-varying treatment using IPTW Norihiro Suzuki1, Shunichiro Orihara2 1School of Data Science, Yokohama City University 2Department of Health Data Science, Tokyo Medical University
因果効果の推定の3ステップ (2/3) ② 各個人への重みの計算 30 /*SW用*/ proc logistic data=df descending; model Trt0=; output out=df predicted=Pr0; run; proc logistic data=df descending; class Trt0/param=ref ref=first; model Trt1=Trt0; output out=df predicted=Pr1; run; /*weightの計算*/ data df_wt; set df; if Trt0=1 then wt0=Pr0/PS0; else wt0=(1-Pr0)/(1-PS0); if Trt1=1 then wt1=Pr1/PS1; else wt1=(1-Pr1)/(1-PS1); wt=wt0*wt1; run;
Reference • 日本製薬工業協会. (2021). リアルワールドデータ・疾患レジストリを活用した医薬品開発. • Robins, J. M., & Hernán, M. A. (2009). Estimation of the causal effects of time-varying exposures. Longitudinal data analysis, 553-599. • Robins, J. M., Hernan, M. A., & Brumback, B. (2000). Marginal structural models and causal inference in epidemiology. Epidemiology, 11(5), 550-560. • Naimi, A. I., Cole, S. R., & Kennedy, E. H. (2017). An introduction to g methods. International journal of epidemiology, 46(2), 756-762. • Kennedy, Edward H. (2019). "Nonparametric causal effects based on incremental propensity score interventions." Journal of the American Statistical Association, 114(526), 645-656. 35