Scribd
Upload
14 views

STAT 709 Midterm Fall 2022

Uploaded by

soumengosh404
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
14 views

STAT 709 Midterm Fall 2022

Uploaded by

soumengosh404
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

STAT 709: Midterm Exam

Instructor: Yiqiao Zhong, TA: Taiyu Ye


Fall 2022

Date: Oct. 25, 2022, 2:30 pm—3:45pm

1. (20 points) Let U ∼ Unif([0, 1]) be a uniform random variable.


p
(1) Prove that U n → 0 as n → ∞.
d
(2) We have convergence in distribution U n → Y . Derive the cumulative distribution function of Y .
(3) Prove that limn→∞ E[U n ] = 0.
(4) Derive the probability density function of U n .

2. (15 points) Recall the last problem from HW1. Let U ∼ Unif([0, 2π]) be a uniform random variable, and
let X, Y ∈ R be given by (
X = cos U,
Y = sin U.

Let µX,Y be the probability measure induced by (X, Y ), and let ν denote the Lebesgue measure1 . Show that
µX,Y  ν and ν  µX,Y are both false.
Hint: You can use the fact that ν({x ∈ R2 : kxk ≤ r}) = πr2 .

3. (10 points) Let X, Y be two independent random variables, and f, g : R → R be two bounded continuous
functions. Use the definition of conditional expectation and independence to prove that

E[f (X)g(Y )|Y ] = E[f (X)]g(Y ) .

4. (15 points) Let (Sn )n≥1 be a sequence of random variables with ESn = 0.
d p
(1) If n−1/2 Sn → N (0, 1), prove that n−1 Sn → 0.
a.s.
(2) If P(n−1 |Sn | ≥ t) ≤ 2 exp(−2nt2 ) holds for every t > 0, prove that n−1 Sn −→ 0.

5. (30 points) True or false questions. Please write T if a statement is correct and F if it is not correct. No
proof of justification is needed.
(1) Let (Ω, F) be a measurable space. If µ satisfies µ(∅) = 0, µ(A) ≥ 0 for any A ∈ F, and µ(∪ni=1 Ai ) =
P n
i=1 µ(Ai ) for any disjoint Ai ∈ F and any positive integer n, then µ is a measure.
(2) A random variable X satisfies E|X| < ∞ if and only if E[max{0, X}] < ∞ and E[max{0, −X}] < ∞.
(3) If a random variable X does not have a (finite) expectation, i.e., E|X| = ∞, then either limx→−∞ FX (x) >
0 or limx→∞ FX (x) < 1, where FX is the c.d.f. of X.
(4) If X, Y, Z are random variables, then E[X|Y, Z] = E[E[X|Y ]|Z]. p
(5)pLet X1 , . . . , Xn be i.i.d. zero mean, subgaussian random variables with kXi kψ2 ≤ 1. Then |Xi | −
E[ |Xi |] is still a subgaussian random variable (so we can use Hoeffding’s inequality to study the concen-
Pn p
tration of n−1 i=1 |Xi |).
1 Strictly speaking, ν is the Lebesgue measure restricted to the Borel σ-algebra B2 .

1
(6) Let ρ ∈ (0, 1) be any constant, n be a positive integer, and [ρn] be the rounding of ρn (to the nearest
integer). If µXn converges weakly a.s. to the semicircular law µsc , then the [ρn]-th largest eigenvalue of a
symmetric random matrix Xn satisfies lim supn λ[ρn] (Xn ) ≤ 2 almost surely.

6. (10 points) Suppose that X1 , . . . Xn ∼i.i.d. N (0, 1), and F : Rn → R is a L-Lipschitz function, i.e.,
|F (x) − F (y)| ≤ Lkx − yk for any x, y ∈ Rn . Prove the following Gaussian concentration inequality: there
exist absolute constants C, c > 0 such that
2
P |F (X1 , . . . , Xn ) − EF (X1 , . . . , Xn )| > Lt ≤ Ce−ct ,

for all t > 0.

d
Hint: let X = (X1 , . . . , Xn )> , and consider an independent copy Y = X. Define Xθ = X cos θ + Y sin θ. Use
the identity Z 2π
d
F (X) − F (Y ) = F (Xθ ) dθ
0 dθ
and Jensen’s inequality.

You might also like