Probability Qualifying Exam Solution, Spring 2011 Shiu-Tang Li December 1, 2012 1 (a) Z e it(X1 +·+XN ) dP = Ω = = ∞ Z X k=1 ∞ X k=1 ∞ X eit(X1 +·+Xk ) dP {N =k} Z P (N = k) eit(X1 +·+Xk ) dP Ω P (N = k)φ(t)k k=1 = φ(t) 2 − φ(t) (b) For each k ∈ N, {N = k} is independent of Sk . 2 (a) Apply L’Hopital’s rule. (b) See Fall 2008 sol, problem 3. 3 See Spring 2010 sol, problem 8. 1 4 Let X1 , X2 , · · · be a sequence of i.i.d random variables with density g(x) = e 1{x>0} . Since E|X1 | < ∞, Strong law of large numbers tell us Snn → E[X1 ] = 1. Since a.s. convergence implies convergence in distribution, we have E[f (Sn /n)] → E[f (1)] = f (1) for every bounded continuous function. f (1) is the answer. −x 5 6 R (a) If X is symmetric, thenR φX (t) = φ−X (t), then Ω i sin tX dP = 0 and R itX hence φX (t) = Ω e dP = Ω cos(tX) dP is real. Conversely, if φX (t) is real, we have φX (t) = φ−X (t). It follows that X and −X has the same distribution function and hence X is symmetric. (b)1. Point mass at 0: φ(t) ≡ 1. 2. Uniform distribution on (−a, a): R Ra 1 a eitx dx = 21 −a cos(tx) dx = sin(at) . 2 −a at 3. N (0, 1) : Z Z (x−it)2 t2 1 − x2 itx 1 − √ e 2 e dx = e 2 √ e− 2 dx 2π 2π R R t2 = e− 2 . Use contour integral to approximate it. 7 If E[X|G ]1G 6= E[Y |G ]1G , then WLOG assume P (E[X|G ]1G > E[Y |G ]1G ) > 0 ⇒ P (E[X|G ]1G > p > q > E[Y |G ]1G ) > 0 for some p, q ∈ Q. Let A = {E[X|G ]1G > p > q > E[Y |G ]1G }. Of course A ∈ G and A ⊂ G. However, by the assumption of the problem we have Z Z E[X|G ] dP = X dP A A Z = Y dP ZA = E[Y |G ] dP A which is a contradiction. 2 8 See also Durrett. E[ρZn |Fn−1 ] = E[ρξ1 +···+ξZn−1 |Fn−1 ]. To compute this, note that Z Z ξ1 +···+ξZn−1 ρ dP = ρξ1 +···+ξk dP Zn−1 =k Zn−1 =k = P (Zn−1 = k)E[ρξ ]k Z k = P (Zn−1 = k)ρ = ρZn−1 . Zn−1 =k 9 E[X 2 ]1/2 E[(1X>a )2 ]1/2 = E[X 2 ]P (X > a) ≥ E[X; X > a] = E[X] − E[X; X ≤ a] ≥ E[X] − a. 10 2 n e−n (1 + n + n2! + · · · + nn! ) = P (Xn ≤ n), where Xn is a Poisson r.v. with parameter n. Let Y1 , Y2 , · · · be a sequence of i.i.d Poisson random variables d of parameter 1, we have Y1 + · · · + Yn = Xn . By central limit theorem, √ n −n ≤ 0) = P (Xn ≤ n) → 1 as n → ∞. P ( Y1 +···+Y 2 n 3