Probability Qualifying Exam Solution, Spring 2008 (SKETCH) Shiu-Tang Li December 27, 2012 1 R 2π R ∞ R∞ Use polar coordinates: 0 0 g(r)r dr dθ = 1. Thus 0 g(r)r dr = 1 . Let U = X/Y , V = Y , we have X = U V , Y = V . fU,V (u, v) = 2π √ √ √ R∞ R∞ R∞ 2 g(|v| u2 + 1)|v|; −∞ g(|v| u2 + 1)|v| dv = 2 0 g(v u2 + 1)v dv = 1+u g(z)z dz = 2 0 1 . π(1+u2 ) 2 Sn −nE[X1 ] (a) CLT: let X1 , · · · , Xn , · · · ∈ L2 (Ω) and be i.i.d RVs, then √ nV ar(X1 ) converges to N (0, 1) weakly. Kolmogrov’s 0-1 law: Let X1 , · · · , Xn , · · · be i.i.d, if A ∈ then P (A) = 0 or 1. T∞ n=1 σ(Xn , Xn+1 , · · · ), (b) P (An i.o.) = P ( ∞ [ ∞ \ Am ) n=1 m=n ∞ [ = lim P ( n→∞ Am ) m=n ≥ inf P (Ak ) > 0. k≥1 (c) We have √ √ √ P (Sn > 0) ≥ P (Sn > log(n)) = P (Sn / n > log(n)/ n) ≥ P (Sn / n > ) 1 for arbitrary > 0 and n = n() large. Therefore by CLT, 1/2 = lim P (Sn > 0) ≥ lim sup P (Sn > log(n)) ≥ lim inf P (Sn > log(n)) n n n √ ≥ lim lim sup P (Sn / n > ) = 1/2. →0+ n As a result, P (limT supn T Sn = S ∞) ≥ limn P (Sn > log(n)) Since T∞ T∞ =S1/2. ∞ ∞ ∞ {lim supn Sn = ∞} = ∞ {S > k} = {S m m− k=1 n=N m=n k=1 n=N m=n SN > k} ∈ σ(XN , XN +1 , · · · , ) for every N ∈ N, it follows that {lim supn Sn = ∞} is a tail event with positive probability, so P (lim supn Sn = ∞) = 1. 3 Consider Ω = {w1 , w2 w3 }, with probability 1/3 each. Let X = 1{w1 ,w2 } , Y = 1{w2 ,w3 } , Z = 1{w1 ,w3 } . 4 This is misstated. Consider X1 ≡ 1, X2 ≡ 3, Xn ≡ 2 for n ≥ 3. 5 Note that X1 /Sn , · · · , Xm /Sn are identically distributed, and the result is straightforward. 6 If k = 1, then P (Tn = k) = 0 for all n. For k > 1, P (Tn = k) = P (X1,n + · · · + Xk,n > n) − P (X1,n + · · · + Xk−1,n > n) = P (X1,n + · · · + Xk−1,n ≤ n) − P (X1,n + · · · + Xk,n ≤ n) n−1 1 n−1 1 − = k nk k − 1 nk−1 1 1 → − as n → ∞. (k − 1)! k! 7 See 2004 problem 7. for relevant techniques. n n 2 t2 1 t2 1 2 EX + o( ) = 1 − + o( ) → e−t /2 . 2n n 2n n 2 n φ( √tn ) = 1 + i √tn EX − 8 P n P (Xn > (1 + ) log(n)) = P n e( (1 + ) log(n)) = 9 10 (Def?) Use inclusion-exclusion principle. 3 P n n1+ .