Exercises Measure Theoretic Probability

Transcription

1 Exercises Measure Theoretic Probability Week 1 1. Prove the folloing statements. (a) The intersection of an arbitrary family of d-systems is again a d- system. (b) The intersection of an arbitrary family of σ-algebras is again a σ- algebra. Characterize σ(c) for a given collection C 2 Ω. (c) If C 1 and C 2 are collections of subsets of Ω ith C 1 C 2, then d(c 1 ) d(c 2 ). 2. Let G and H be to σ-algebras on Ω. Let C = {G H : G G, H H}. Sho that C is a π-system and that σ(c) = σ(g, H). 3. Sho that D 2 (Williams, page 194) is a π-system. 4. If h 1 and h 2 are measurable functions, then h 1 h 2 is measurable too. 5. Let Ω be a countable set. Let F = 2 Ω and let p : Ω [0, 1] satisfy ω Ω p(ω) = 1. Put P(A) = ω A p(ω) for A F. Sho that P is a probability measure. 6. Let Ω be a countable set. Let A be the collection of A Ω such that A or its complement has finite cardinality. Sho that A is an algebra. What is d(a)? Week 2 1. Let X be a random variable. Sho that Π(X) := {X 1 (, x] : x R} is a π-system and that it generates σ(x). 2. Let {Y γ : γ C} be an arbitrary collection of random variables and {X n : n N} be a countable collection of random variables, all defined on the same probability space. (a) Sho that σ{y γ : γ C} = σ{yγ 1 (B) : γ C, B B}. (b) Let X n = σ{x 1,..., X n } (n N) and A = n=1 X n. Sho that A is an algebra and that σ(a) = σ{x n : n N}. 3. Sho that the X + and X are measurable functions and that X + is right-continuous and X is left-continuous (notation as in section 3.12). 4. Williams, exercise E4.1. 1

2 5. Williams, exercise E Let F be a σ-algebra on Ω ith the property that for all F F it holds that P(F ) {0, 1}. Let X : Ω R be F-measurable. Sho that for some c R one has P(X = c) = 1. (Hint: P(X x) {0, 1} for all x.) Week 3 1. Let X and Y be simple random variables. Sho that E X doesn t depend on the chosen representation of X. Sho also that E (X +Y ) = E X +E Y. 2. Sho that the expectation is a linear operator on L 1 (Ω, F, P). 3. Let (Ω, F, P) = ([0, 1], B, Leb) and let f, f 1, f 2,... be densities (nonnegative measurable functions that integrate to 1) on [0, 1]. Assume that f n f a.s. Sho that for all x [0, 1] it holds that F n (x) F (x), here F n (x) = [0,x] f n(t) dt and F (x) = f(t) dx. [0,t] 4. (a) Sho that for X L 1 (Ω, F, P) it holds that E X E X. (b) Prove the second part of Scheffé s lemma for random variables X and X n (see page 55). 5. Let X L 1 (Ω, F, P). Sho that lim n np( X > n) = 0. Week 4 1. Prove the assertions (a)-(d) of section Prove the assertions (a)-(c) of section Prove lemma 6.12 (you find the standard machine in section 5.12). 4. Prove part (b) of Fubini s theorem in section Complete the proof of theorem 6.11: sho the a.s. uniqueness of Y and sho that if X Y Z for all Z in K, then X Y 2 = inf{ X Y 2 : Y K}. Week 5 1. Finish the proof of theorem 9.2: Take arbitrary X L 1 (Ω, F, P) and sho that the existence of the conditional expectation of X follos from the existence of the conditional expectations of X + and X. 2. Prove the conditional version of Fatou s lemma, property (f) on page Prove the conditional Dominated Convergence theorem, property (g) on page 88. 2

3 4. Exercise E Exercise E Let (X, Y ) have a bivariate normal distribution ith E X = µ X, E Y = µ Y, Var X = σx 2, Var Y = σ2 Y and Cov (X, Y ) = c. Let ˆX = µ x + c σ 2 Y (Y µ Y ). Sho that E (X ˆX)Y = 0. Sho also (use a special property of the bivariate normal distribution) that E (X ˆX)g(Y ) = 0 if g is a Borelmeasurable function such that E g(y ) 2 <. Conclude that ˆX is a version of E [X Y ]. Week 6 In all exercises belo e consider a probability space (Ω, F, P) ith a filtration F. 1. Let X be an adapted process and T a stopping time that is finite. Sho that X T is F-measurable. Sho also that for arbitrary stopping times T (so the value infinity is also alloed) the stopped process X T is adapted. 2. Sho that an adapted process X is a martingale iff E[X n+m F n ] = X n for all n, m Read first the definition of a submartingale. Let f be a convex function on R. Suppose that X is a martingale and that E f(x n ) < for all n. Sho that the process (f(x n )) n 0 is a submartingale. 4. For every n e have a measurable function f n on R n. Let Z 1, Z 2,... be independent random variables and F n = σ(z 1,..., Z n ). Sho that (you may assume sufficient integrability) that X n = f n (Z 1,..., Z n ) defines a martingale under the condition that Ef n (z 1,..., z n 1, Z n ) = f n 1 (z 1,..., z n 1 ) for every n. 5. If S and T are stopping times, then also S + T, S T and S T are stopping times. Sho this. Week 7 In all exercises belo e consider a probability space (Ω, F, P) ith a filtration F. 1. (a) If X is a martingale is and f a convex function such that E f(x n ) <, then Y defined by Y n = f(x n ) is a submartingale. Sho this (keyord: Jensen). 3

4 (b) Sho that Y is a submartingale, if X is one and f is an increasing function. 2. Prove Corollaries (c) and (d) on page Sho that the process C of section 11.1 is previsible. 4. Let X be a submartingale ith sup n 0 E X n <. Sho that there exists a random variable X such that X n X a.s. 5. Sho that for a supermartingale X the condition sup{e X n : n N} < is equivalent to the condition sup{e X n : n N} <. 6. Consider the probability space (Ω, F, P) ith Ω = [0, 1), F the Borelsets of [0, 1) and P the Lebesgue measure. Let Ik n = [k2 n, (k + 1)2 n ) for k = 0,..., 2 n 1 and F n be the σ-algebra by the Ik n for k = 0,..., 2n 1. Define X n = 1 I n 0 2 n. Sho that X n is a martingale is and that the conditions of theorem 11.5 are satisfied. What is X in this case? Do e have L X 1 n X? (This has something to do ith 11.6). Week 8 1. Let Y L 1, (F n ) and define for all n N the random variable X n = E [Y F n ]. We kno that there is X such that X n X a.s. Sho that for Y L 2 L, e have X 2 n X. Find a condition such that X = Y. Give also an example in hich P (X = Y ) = Sho that every finite collection in L 1 is uniformly integrable. 3. Williams, exercise E Let C be a uniformly integrable collection of random variables. (a) Consider C, the closure of C in L 1. Use E13.1 to sho that also C is uniformly integrable. (b) Let D be the convex hull of C. Then both D and its closure in L 1 are uniformly integrable 5. In this exercise you prove (fill in the details) the folloing characterization: a collection C is uniformly integrable iff there exists a function G : R + R + G(t) such that lim t t = and M := sup{eg( X ) : X C} <. The necessity you prove as follos. Let ε > 0 choose a = M/ε and c such a for all t > c. To prove uniform integrability of C you use that X G( X ) a on the set { X c}. It is less easy to prove sufficieny. Proceed as follos. Suppose that e have a sequence (g n ) ith g 0 = 0 and lim n g n =. Define g(t) = that G(t) t n=0 1 [n,n+1)(t)g n and G(t) = t 0 g(s)ds. Check that lim t G(t) t =. 4

5 With a n (X) = P( X > n), it holds that E G( X ) n=1 g na n ( X ). Furthermore, for every k N e have X k X dp m=k a m(x). Pick for every n a constant c n N such that X c n X dp 2 n. Then m=c n a m (X) 2 n and hence n=1 m=c n a m (X) 1. Choose then the sequence (g n ) as the inverse of (c n ): g n = #{k : c k n}. 6. Prove that a collection C is uniformly integrable iff there exists an increasing and convex function G : R + R + G(t) such that lim t t = and M := sup{g( X ) : X C} <. Let D be the closure of the convex hull of a uniformly integrable collection C in L 1. With the function G as above e have sup{eg( X ) : X D} = M, hence also D is unifromly integrable. 7. Let p 1 and let X, X 1, X 2,... be random variables. Then X n converges to X in L p iff the folloing to conditions are satisfied. Week 9 (a) X n X in probability, (b) The collection { X n p : n N} is uniformly integrable. 1. Sho that R, defined on page 142 of Williams, is equal to qe X p 1 Y. Sho also that the hypothesis of lemma is true for X n if it is true for X and complete the proof of this lemma. 2. Exercise E14.2 of Williams. 3. Let X = (X n ) n 0 a (backard) supermartingale. (a) Sho equivalence of the next to properties: (i) sup n E X n < and (ii) lim n EX n <. (Use that x x + is convex and increasing.) (b) Under the condition sup n E X n =: A < the supermartingale X is uniformly integrable. To sho this, you may proceed as follos (but other solutions are equally elcome). Let ε > 0 and choose K Z such that for all n < K one has 0 EX n EX K < ε. It is then sufficient to sho that (X n ) n K is uniformly integrable. Let c > 0 be arbitrary and F n = { X n > c}. Using the supermartingale inequality you sho that X n dp X K dp + ε. F n F n Because P(F n ) A c you conclude the proof. 5

6 Week Let µ, µ 1 µ 2,... be probability measures on R. Sho that µ n µ iff for all bounded Lipschitz continuous functions one has f dµ n f dµ. (Hint: for one implication the proof of lemma 17.2 is instructive.) 2. Sho the if part of lemma If the random variables X, X 1, X 2,... are defined on the same probability P space and if X n X, then Xn X. Prove this. 4. Suppose that X n X and that the collection {Xn, n 1} is uniformly integrable (you make a minor change in the definition of this notion if the X n are defined on different probability spaces). Use the Skorohod representation to sho that X n X implies EX n EX. 5. Sho the folloing variation on Fatou s lemma: if X n X, then E X lim inf n E X n. 6. Sho that the eak limit of a sequence of probability measures is unique. Week Consider the N(µ n, σ 2 n) distributions, here the µ n are real numbers and the σ 2 n nonnegative. Sho that this family is tight iff the sequences (µ n ) and (σ 2 n) are bounded. Under hat condition do e have that the N(µ n, σ 2 n) distributions converge to a (eak) limit? What is this limit? 2. For each n e have a sequence ξ n1,..., ξ nkn of independent random variables ith Eξ nj = 0 and k n j=1 Var ξ nj = 1. If k n j=1 E ξ nj 2+δ 0 as n for some δ > 0, then k n j=1 ξ nj N(0, 1). Sho that this follos from the Lindeberg Central Limit Theorem. 3. The classical central limit theorem says that 1 σ n n j=1 (X j µ) N(0, 1), if the X j are iid ith EX j = µ and 0 < Var X j = σ 2 <. Sho that this follos from the Lindeberg Central Limit Theorem. 4. Sho that X n X iff Ef(X n ) Ef(X) for all bounded uniformly continuous functions f. 5. Let X and Y be independent, assume that Y has a N(0, 1) distribution. Let σ > 0. Let φ be the characteristic function of X: φ(u) = E exp(iux). (a) Sho that Z = X+σY has density p(z) = 1 σ 2π E exp( 1 2σ (z X) 2 ). 2 (b) Sho that p(z) = 1 2πσ φ( y/σ) exp(iyz/σ 1 2 y2 ) dy. 6

7 6. Let X, X 1, X 2,... be a sequence of random variables and Y a N(0, 1)- distributed random variable independent of that sequence. Let φ n be the characteristic function of X n and φ that of X. Let p n be the density of X n + σy and p the density of X + σy. (a) If φ n φ pointise, then p n p pointise. Invoke the previous exercise and the dominated convergence theorem to sho this. (b) Let f C b (R) be bounded by B. Sho that Ef(X n +σy ) Ef(X + σy ) 2B (p(z) p n (z)) + dz. (c) Sho that Ef(X n + σy ) Ef(X + σy ) 0 if φ n φ pointise. (d) Prove the folloing theorem: X n Week 12 X iff φn φ pointise. 1. Consider the sequence of tents (X n ), here Xt n = nt for t [0, 1 2n ], Xt n = 1 nt for t [ 1 2n, 1 n ], and zero elsehere (there is no randomness here). Sho that all finite dimensional distributions of the X n converge, but X n does not converge in distribution. 2. Sho that ρ as in (1.1) defines a metric. 3. Suppose that the ξ i of section 4 are iid normally distributed random variables. Use Doob s inequality to obtain P(max j n S j > γ) 3γ 4 n Sho that a finite dimensional projection on C[0, ) (ith the metric ρ) is continuous. 5. Consider C[0, ) ith the Borel σ-algebra B induced by ρ and some probability space (Ω, F, P). If X : (Ω, F) (C[0, ), B) is measurable, then all maps ω X t (ω) are random variables. Sho this, as ell as its converse. For the latter you need separability that allos you to say that the Borel σ-algebra B is a product σ-algebra (see also Williams, page 82). 6. Prove proposition 2.2 of the lecture notes. 7