e) = f(s − e) ∫ T −e 0 re−ru 1 − e−ru F(T − e − u)du ͜͜Ͱ, r ࢦؔత૿ՃͰ F(·) f(·) ͷྦྷੵؔ. T ࠷৽ͷ ؍ଌ࣌ࠁ3Λࣔ͢. ͯ͞, ैͬͯҎԼͷΑ͏ʹͳΔ. ·ͨ͜ͷͱ͖ΠϯςάϥϧҎԼΛ ιi ͱͯ࣍͠ͷΑ͏ʹද͢. L′(Θ|D) = ∏ i ∫ ER,i EL,i ∫ SR,i SL,i g(e)f′(s − e, e)dsde ιi = ∫ ER,i EL,i ∫ SR,i SL,i g(e)f′(s − e, e)dsde (3.1) SL,i > ER,i ͷͱ͖, s′ = s − e ͱஔͯ࣍͠ͷΑ͏ʹมܗͰ͖Δ. ιi = ∫ ER,i EL,i de g(e) ∫ SR,i−e SL,i−e f′(s′, e)ds′ = ∫ ER,i EL,i g(e) {F′(SR,i − e, e) − F′(SL,i − e, e)} de (3.2) ER,i > SL,i > EL,i Ͱ͋Δͱ͖, ҎԼͷΑ͏ʹͳΔ. ιi = ∫ Sl,i EL,i g(e){F′(SR,i − e, e) − F′(SL,i − e, e)}de + ∫ ER,i SL,i g(e)F′(SR,i − e, e)de (3.3) ࠷ޙʹ, EL,i > SL,i Ͱ͋Δͱ͖, ҎԼͷΑ͏ʹͳΔ. ιi = ∫ ER,i EL,i g(e)F′(SR,i − e, e)de (3.4) ͜ͷΑ͏ʹ߹͚Λͯ͠ time interval ΛͦΕͧΕٻΊͨ.4 f(s − e) Ψ ϯϚ, ରਖ਼ن, ϫΠϒϧΛߟ͑ͨ. 1ੜଘ࣌ؒղੳͷͰ͢. ଧͪΓ censoring ͱஅ truncation ͕ߟྀ͞ΕΔ͖ิਖ਼Ͱ ͢. ࠓճͷ right truncation જ෬ظ͕͍ؒ߹ʹΤϯυϙΠϯτͷઃఆʹΑΓ؍ଌ͞Εͣ, population of interest ͱ sampled population ͕ҟͳͬͯ͠·͍ selection bias ͕ൃੜ͢Δঢ় ଶΛ͍͏. Inverse probability weighting methods for Cox regression with rightʖtruncated data (https://doi.org/10.1111/biom.13162). 2Selection bias ͷิਖ਼ͷํΑ͘Θ͔Βͳ͍Ͱ͢... 3͜ͷͱ͖ 2020 1 ݄ 31 . 4࣮ࡍͷσʔλΛ༻͍ͨܭࢉ R Λ༻ͨ͠. 8