关键词 > STAT3004/STAT4018/STAT6018

STAT3004/STAT4018/STAT6018 Questions

发布时间:2023-06-08

Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

STAT3004/STAT4018/STAT6018

Question 1                                   2+2+2+2+2=10 pts

Without giving reasons, choose the answer which best completes each statement below. Each correct answer will gain two marks, there is no penalty for incorrect answers.

(a) The significance of σ -algebras is best explained in the context of

(1) the carrier of information;

(2) discrete random variables;

(3) the domain of probability measures and the carrier of information;

(4) Laplace experiments.

(b) A function f : R - R is a Borel function

(1) always;

(2) provided f is an indicator function;

(3) provided f is nowhere continuous;

(4) provided f is monotone.

(c) A set system is a generator of the Borel σ -algebra on the real line provided

(1) it is closed under taking nite intersections;

(2) it is closed under taking uncountable unions;

(3) it contains all the singletons;

(4) it contains all the intervals.

(d) If E[X] = 6 and E[X2] = 40, then Chebychev’s inequality states

(1) P(|X _ 6| > 4) < 0.125;

(2) P(|X _ 6| > 4) < 0.1111;

(3) P(|X _ 6| > 4) < 0.5;

(4) P(|X _ 6| > 4) < 0.25.

(e) Assume r is a σ -algebra, defined on a non-empty set while µ : r - [0, o],

F -l µ(F), is a given function.

(1) If µ(Ω) = 2, then µ is a nite measure;

(2) If µ is countably additive, then µ must be a probability measure;

(3) The Lebesgue measure µ is a σ -finite measure when defined on r = B(R);

(4) The counting measure µ is a probability measure.

Question 2 2+2+2+2+2=10 pts

Without giving reasons, choose the answer which best completes each statement below. Each correct answer will gain two marks, there is no penalty for incorrect answers.

(a) If X is a real-valued random variable with nite E[|X|], then

(1) E[X3] is well-defined provided X is standard Cauchy;

(2) E[X] is well-defined and nite;

(3) E[X] could be infinite, while still well-defined;

(4) E[X3] is well-defined provided X is symmetric.

(b) If t -l ψX (t) is the characteristic function of X ,

(1) ψ may not be continuous at the origin;

(2) _iψ\ (0) is the expected value, provided the rst moment exists;

(3) _ψ \ (0) is the expected value, provided the rst moment exists;

(4) t -l e_i2t  is the characteristic function of the standard normal distribution.

(c) If λ is Lebesgue measure and (Un )n  is a sequences of independent random vari- ables, uniformly distributed in the unit interval [0, 1], the law of

(1) 2_nI[0,1/2)(Un ) is not absolutely continuous with respect to λ;

(2)     n(4)=1 2_nI[0,1/2)(Un ) is absolutely continuous with respect to λ;

(3) 2 . 3_nI[1/2,1] (Un ) is not absolutely continuous with respect to λ;

(4) 2_nI[0,1/3)(U15 ) is not a discrete random variable.

(d) Assume X, X1 , X2 . . .  is a sequence of real-valued random variables, defined on

a common probability space  (Ω , r, P), where, for n e N , Xn  := k(n)=1 Xk   is

the associated sample mean.

(1) if Xn  - X in probability as n - o , then Xn  - X almost surely as n - o ;

(2) if E[Xn]=2 and Var(Xn )=3 for all n e N, then Xn  - 2 as n - o ;

(3) if E[X2] + E[Xk(2)] < o , then Xn - 0 almost surely;

(4) if E[X2]+    k(n)=1 E[(Xk _ X)2] < o for n e N, then limno E[(Xn _ X)2]=0 .

(e) Assume X, Y are a random variables, defined on some probability space (Ω , r, P),

so that E[|X|] + E[|Y |] < o . Assume s , g C r are σ -algebras.

(1) if s C g and s g , then E[X|s] < E[X|g], P-a.s.;

(2) if s C g and σ(Y) = g , then E[XY |s] = XE[Y |s], P-a.s.;

(3) always, E[Y |g] = E[Y |σ(s u g)], P-a.s.;

(4) if s C g and σ(Y) = g , then E[XY |g] = YE[X|g], P-a.s..

Here P-a.s. indicates that, with probability one, an identity or (strict) inequality holds for the associated versions.

Question 3                                3+4+3+5=15 pts

(a) If X is a random variable dened on a probability space (Ω , r, P), by using the

denition of measurability, show that |X| is a random variable.

(b) By an example, show that the converse of (a) need not to be true. That is give

an example such that |X| is a random variable but X is not a random variable.

(c) Show thatE[X] < E [|X|] .

(d) If g C r is a sub σ -algebra, show that the modulus inequality holds for the conditional expectation, i.e. E[X|g] < E [|X| g] with probability one.

Question 4                         3+4+4+4=15 pts

(a) Use the exponential series ez = exp(z) = zj /j!, z = x+iy e C , x, y e R to

determine the characteristic function of the Poisson(2) distribution.

(b) The normal distribution Normal(m, v) with mean m e R and variance v > 0 has Lebesgue density exp(_(y _m)2 /(2v))/)2πv , y e R, and Fourier transformation exp(imt _ vt2 ), t e R .

Assume Y = 5Y1 +2Y2 _ 2Y3 where Y1 , Y2 , Y3 are independent random variables with 1(Y1 ) = 1(Y2 ) = 1(Y3 ) =Normal(1, 2) .

Find (i) the characteristic function and (ii) the Lebesgue density of Y .

(c) Verify that a random variable W is symmetrically distributed, i.e.   1(W)  = 1(_W), if and only if its characteristic function is real-valued.

(d) Assume Z1 , Z2 , . . .  is a sequence of independent and identically distributed ran- dom variables with characteristic function ψZ1 (t)  = e_3ltl , t e R . Prove or disprove that the family {n Zj : n e N} is tight.

Question 5 3+4+4+4=15 pts

Let Q be the set of rational numbers.  Let B = B(R) = σ({(_o, x] : x e R}) be the Borel field. Assume P is a probability measure on B . Prove or disprove

(a) if B1 , B2 , . . . e B so that P(Bj ) = o , then P(lim infno Bn ) = 1/2; (b)  B = σ({(_o, q] : q e Q});

(c)  B = σ({{q} : q e Q});

(d)  B = σ(Fb ) where Fb  is the set of bounded functions from R to R .

Question 6                    8+7=15 pts

Let U1 , U2 , . . . be a sequence of independent and uniformly distributed random vari- ables on the interval [0, θ] . We interpret θ > 0 as an unknown parameter.

Let  Un   = k(n)=1 Uk   and  Mn   = max1

(a)  Show that

(i) 2Un  - θ almost surely as n - o ;

(ii) Mn  - θ almost surely as n - o .

(b) Determine the weak limits of their error distributions, respectively. Which esti-mator is preferable?