Memoryless Property βΉ Geometric Distribution Show that if a random variable π deο¬ned on {1, 2, 3, …} satisο¬es β(π = π + π | π ≥ π + 1) = β(π = π), π = 1, 2, 3, … for every integer π ≥ 1, then π follows a geometric(π) distribution. Setting π = 1, we have for π = 1, 2, 3, … β(π = π + 1 | π ≥ 2) = β(π = π). β(π = π + 1) = β(π = π + 1 ∩ π ≥ 2) since {π = π + 1} ⊂ {π ≥ 2} = β(π ≥ 2) × β(π = π + 1 | π ≥ 2) = π × β(π = π) by deο¬nition of conditional probability by (1) where π βΆ= β(π ≥ 2). It follows by induction that β(π = π) = πuοΏ½−1 π where π βΆ= β(π = 1) ≠ 0 because otherwise β is identically zero. (1) Memoryless Property βΉ Exponential Distribution Show that if a continuous random variable π deο¬ned on β≥0 satisο¬es β(π > π‘0 + π‘ | π > π‘0 ) = β(π > π‘), for every π‘0 ≥ 0, then π follows an exponential(π) distribution. π‘≥0 β(π > π‘) × β(π > π‘0 ) = β(π > π‘0 + π‘ | π > π‘0 ) × β(π > π‘0 ) by given equality = β(π > π‘0 + π‘ ∩ π > π‘0 ) by deο¬nition of conditional probability = β(π > π‘0 + π‘) since {π > π‘0 + π‘} ⊂ {π > π‘0 }. Deο¬ne the survival function πΊ(π₯) = β(π > π₯). Then we have the semigroup property πΊ(π‘ + π‘0 ) = πΊ(π‘) ⋅ πΊ(π‘0 ). We see that πΊ(0) = 1 by setting π‘0 = 0. (2) πΊ must be diο¬erentiable by continuity of π and the Fundamental Theorem of Calculus, so we can ο¬x π‘ and diο¬erentiate both sides of (2) w.r.t π‘0 using the Chain Rule: In particular, setting π‘0 = 0 gives πΊ′ (π‘ + π‘0 ) = πΊ(π‘) ⋅ πΊ′ (π‘0 ) πΊ′ (π‘) = πΊ(π‘) ⋅ πΊ′ (0). Therefore, letting π = −πΊ′ (0), we have πΊ′ (π‘) = −ππΊ(π‘), which has solution πΊ(π‘) = π−uοΏ½uοΏ½ as required. πΊ(0) = 1 Since πΊ is a decreasing function, we must have π > 0.