Math 630 Notes Survival Models • Furure Life Time Tx and Its Distribution – (x): a life aged x – Tx : the future life time of (x) Example: Consider a life (55). If T55 = 30, then (55) dies at age 55 + T55 = 85. – The CDF of Tx : Fx (t) = P (Tx ≤ t), the probability that (x) dies within t years – Survival function of (x): Sx (t) = 1 − Fx (t) = P (Tx > t), the probability that (x) survives for at least t years – Notation: t px = Sx (t) = P (Tx > t), t qx = Fx (t) = P (Tx ≤ t). For t = 1, we drop the front subscript and use px = 1 px and qx = 1 qx – An important formula P (Tx ≤ t) = P (T0 ≤ x + t|T0 > x). – Similarly, P (Tx > t) = P (T0 > x + t | T0 > x) – Exercise Show that 1) Fx (t) = F0 (x + t) − F0 (x) S0 (x) 2) Sx (t) = S0 (x + t) or S0 (x + t) = S0 (x)Sx (t) S0 (x) 3) Sx (t + u) = Sx (t)Sx+t (u) – Conditions on Sx (t): Sx (0) = 1, lim Sx (t) = 0, Sx (t) is non-increasing. t→∞ lim t2 Sx (t) = 0 ( ⇒ lim tSx (t) = 0.) t→∞ √ 1 – Example (Exercise 2.3) Given the survival function S0 (x) = 10 100 − x for 0 ≤ x ≤ 100, find the probability that (0) will die between ages 19 and 36. – Assumptions: Sx (t) is smooth, t→∞ – Example. You are given the following survival data of a group of 100 people, where lx is the number of people in the group who survive to age x. x lx 0 100 1 97 2 ··· 94 · · · 50 32 51 28 52 26 53 23 Find F0 (53), S0 (51), f0 (51), S2 (50), and the probability that someone age 2 will die between 51 and 53. P (51 < T0 ≤ 53 | T0 > 2) = S0 (51) − S0 (53) = P (49 < T2 ≤ 51) S0 (2) or just count how many among the 94 survived to age 2 die between 51 and 53. • The Force of Mortality 1 1 1 P (Tx ≤ ∆x) = lim P (T0 ≤ x + ∆x | T0 > x) = lim Fx (∆x) – µx = lim ∆x→0 ∆x ∆x→0 ∆x ∆x→0 ∆x – µx ∆x ≈ P (Tx ≤ ∆x) S0 (x + ∆x) S0 (x) − S0 (x + ∆x) = implies that S0 (x) S0 (x) 1 S0 (x) − S0 (x + ∆x) −1 dS0 (x) −S00 (x) µx = lim = = ∆x→0 ∆x S0 (x) S0 (x) dx S0 (x) – Fx (∆x) = 1 − Sx (∆x) = 1 − −S00 (x) d = − ln S0 (x) S0 (x) dx – Let f0 (x) be the density function of T0 . Then −S00 (x) = F00 (x) = f0 (x), and so – µx = – µx = f0 (x) S0 (x) – Let x be fixed and t be variable. Then µx+t = · · · = – µx+t = −Sx0 (t) −1 d Sx (t) = Sx (t) dt Sx (t) −Sx0 (t) fx (t) d = = − ln Sx (t) Sx (t) Sx (t) dt d – Integrating µx+s = − ds ln Sx (s) from s = 0 to s = t, we get Z t Z Sx (t) = exp − µx+s ds = exp − 0 x+t µr dr . x – µx ⇐⇒ Sx – Example. (Exercise 2.1 (a)–(d)) Let F0 (t) = 1 − (1 − t/105)1/5 for 0 ≤ t ≤ 105. Calculate (a) the probability that a newborn dies before age 60. (b) the probability that a life aged 30 survives to at least age 70. (b) the probability that a life aged 20 dies between ages 90 and 100. (a) the force of mortality at age 50. – Example. (Exercise 2.5 (a) and (b)) Let F0 (t) = 1 − e−λt , where λ > 0. (a) Show that Sx (t) = e−λt . (b) Show that µx = λ. – Special Mortality Laws ∗ ∗ ∗ ∗ Constant mortality: µx = c Gompertz’ Law: µx = Bcx (See Example 2.3) Makeham’s Law: µx = A + Bcx 1 De Moivre’s Law: µx = ω−x for 0 ≤ x < ω – Example Given that µx is constant µ and that the probability that a life aged 60 survives to age 80 is 0.1, find µx . Z 20 µx+t dt = 0.1 ⇒ e−20µ = 0.1 ⇒ µ = 0.11513. 20 p60 = 0.1 ⇒ exp − 0 – Example. DML with ω = 100. 1 µx = for 0 ≤ x < 100 ⇒ for 0 ≤ t < 100 − x, 100 − x Z t 1 100 − (x + t) ds = , t px = exp − 100 − x 0 100 − (x + s) t qx = t . 100 − x (Use a time line to express these expressions.) • More on Notations – t px = Sx (t) = P (Tx > t), t qx = Fx (t) = P (Tx ≤ t), u|t qx = P (u < Tx ≤ u + t) = Sx (u) − Sx (u + t) = Fx (u + t) − Fx (u) = u pxt · qx+u = u px − u+t px = u+t qx − u qx , u+t px = u px · t px+u 1 d 1 d – µx = − (x p0 ), µx+t = − (t px ) x p0 dx t px dt Z t d – fx (t) = Fx (t) = t px µx+t , t qx = s px µx+s ds dt 0 – u|t qx • Mean and Variance of Tx Z ∞ Z ∞ Z ∞ ◦ tt px µx+t dt = · · · = tfx (t) dt = – ex = E(Tx ) = t px dt 0 0 0 Z ∞ Z ∞ 2 2 – E(Tx ) = t fx (t) dt = · · · = 2 t · t px dt 0 V(Tx ) = E(Tx2 ) 0 2 − [E(Tx )] = E(Tx2 ) ◦ 2 − ex – Example For constant force of mortality µx = 0.3, t px = e−0.3t ⇒ Z ∞ 1 1 ◦ ex = = . e−0.3t dt = 0.3 µx 0 Z ∞ 2 2 E(Tx ) = 2 t · e−0.3t dt = · · · = ⇒ 0.32 0 2 2 1 1 V(Tx ) = − = . 2 0.3 0.3 0, 32 In general, for constant µx = µ, ◦ ex = E(Tx ) = 1 , µ V(Tx ) = 1 . µ2 – Example. (DML with ω = 100) 100−x Z ◦ ex = 0 t px = 100−(x+t) 100−x ⇒ 100 − (x + t) 100 − x dt = . 100 − x 2 It can be computed that V(Tx ) = (100 − x)2 . 12 In general ◦ ex = ω−x , 2 V(Tx ) = (ω − x)2 . 12 • Curtate Life Time Kx – Definition: Kx = bTx c, the integer part of Tx . – P (Kx = k) = P (k ≤ Tx < k + 1) = k| qx = k px − k+1 px = k px · qx+k – ex = E(Kx ) = ∞ X kP (Kx = k) = k=1 – Similarly, E(Kx2 ) V(Kx ) = 2 ∞ X ∞ X k(k px − k+1 px ) = · · · = k=1 = ··· = 2 ∞ X k · k px − k=1 ∞ X k px k=1 ∞ X k px =2 ∞ X k · k px − ex and so k=1 k=1 k · k px − ex − e2x k=1 – Example. Find e50 for DML with ω = 100. – Exercise. Find e50 for CFM with µx = 0.3. ◦ ◦ – Relation between ex and ex : ex ≈ ex + 21 . • Suggested Exrecises – From the text: Exercises 2.1–2.3, 2.6–2.7, 2.10–2.11 (Exercises 2.9, 2.14, and 2.15 require more calculus; they are strongly recommended.) 100 2 . – Find 5| q40 if S0 (t) = 100+t – Given µx = 2 100−x for 0 ≤ x < 100. Calculate 10| q65 . – Under DML with ω = 100, calculate the probability that (30) will die in his 70’s. ◦ 0.01 if t ≤ 5 – Given that µ70+t = , calculate e70 . 0.02 if t > 5 (Hint: Find explicit expressions for t p70 for t ≤ 5 and for t > 5.) R∞ ◦ Appendix It is shown in the text that ex = 0 t px dt using integration by parts. In order to use integration by parts, it is assumed that lim tSx (t) = 0. Here is an alternative proof t→∞ without making the further assumption. First a general resulty in probability. Lemma If X is a non-negative continuous random variable with CDF F (x) and if E(X) exists, then Z ∞ E(X) = (1 − F (t)) dt. (1) 0 Proof. Let f (x) = F 0 (x), the density function of X. Then, we have Z ∞ Z s Z ∞ dt · f (s) ds s · f (s) ds = E(X) = 0 Z0 ∞ Z s Z ∞0Z ∞ = f (s) dt ds = f (s) ds dt 0 0 0 t Z ∞Z ∞ Z ∞ = f (s) ds dt = P (X > t) dt 0 t 0 Z ∞ = (1 − F (t)) dt 0 Now for X = Tx , F (t) = P (Tx ≤ t) = t qx , and 1 − F (t) = 1 − t qx = t px . We have by (1) Z ∞ Z ∞ ◦ ex = E[Tx ] = (1 − F (t)) dt = t px dt. 0 0 The formula for E(Tx2 ) and V(Tx ) can be similarly onbtained.
© Copyright 2026 Paperzz