Professional Documents
Culture Documents
Spring 2016
Problem Set 2
Author: Xiaobin Gao
1. Problem 1
p.
Yes, Xn 1.
Proof. Given that lim P (An ) = 1, for any > 0,
n
p.
By definition, Xn 1.
2. Problem 2
No, Xn2 does not necessarily converge in mean square.
Proof. Let {Xn , n 1} be random variables defined on standard unit-interval probability
space. Furthermore,
n1/4 , if 0 1
n
Xn () =
0,
1
if n < 1
Then
E[(Xn 0)2 ] =
1
n
n
m.s.
Xn
Moreover,
E[Xn4 ] = 1 lim E[Xn4 ] = 1 <
n
n1/2 , if 0
0,
if
1
n
1
n
<1
a.s.
It is easy to see that Xn 0. Then if {Xn } converges in the mean square sense, {Xn }
p.
p.
a.s.
m.s.
converges to 0. (Since Xn 0, Xn 0. Suppose Xn X, then Xn X, which implies
P (X = 0) = 1) However,
E[(Xn2 0)2 ] = E[Xn4 ] = 1 lim E[(Xn2 0)2 ] = 1 6= 0
n
3. Problem 3
Yes, {Xn } converges almost surely to 0.
Proof. For any realization of X, denoted by x, Xn (x) = xn , x [1, 1]. Then,
0,
if 1 < x < 1
lim Xn (x) = 1,
if x = 1
n
not exist , if x = 1
which implies
P {x : lim Xn (x) = 0} = P (1 < x < 1) = 1
n
a.s.
By definition, Xn 0.
4. Problem 4
d.
Yes, Xn + an X + a.
d.
Proof. Since Xn X,
lim E ejuXn = E ejuX , u R
Then,
h
i
h
i
lim E eju(Xn +an ) = lim ejuan lim E ejuXn = ejua E ejuX = E eju(X+a) , u R
d.
(a) Yes, Xn X.
Proof. Define
n := (1)n X =
X
X,
if n is even
X, if n is odd
Since X has uniform distribution over [1, 1], it is easy to see that X also has uniform
distribution over [1, 1]. Hence FXn (x) = FX (x), n 1, x R, which implies that
d.
n
X
X. Furthermore, lim 1 = 0. By results from Problem 4, we conclude that
n n
d.
Xn X.
p.
m,n
Without loss of generality we can assume m is even, n is odd, and m < n (hence
Consider = 0.2 > 0,
1
m
> n1 ).
1
1
1
P (|Xm Xn | ) = P 2X +
P X = 0.45
m n
2
Then,
lim P (|Xm Xn | ) 0.45 6= 0
m,n
6. Problem 6
(a) Yes, {Xn } converges almost surely.
Proof. {Xn , n 1} is a sequence of random variables defined on the probability space
(, F, P ). The sample space , = [0, 1] , is the collection of all realizations of random
sequence {Un , n 1}. Consider any , it is easy to see that
0 Xn () Xn+1 () 1, n 1
Since {Xn (), n 1} is a bounded monotone increasing sequence, it converges. By
Cauchy criterion for deterministic sequence,
lim |Xn () Xm ()| = 0
m,n
Hence,
P :
lim |Xn () Xm ()| = 0 = 1
m,n
a.s.
(c) {Xn } 1.
E[Xn ] =
=
Xn1 ,Un
[Xn ]
E [Xn |Xn1 ]
Xn1 Un
"Z
=
Xn1
Xn1
Xn1
#
Xn1 + u
fUn (u)du
2
1 2
1
1
Xn1 + Xn1 +
4
2
4
2 1
1
1
E[Xn1 ] + E[Xn1 ] +
4
2
4
E
Xn1
where fUn (u) = 1{0u1} is the pdf of Un . The last inequality dues to Jensens inequality.
Since {Xn } converges in the mean square sense, {Xn } converges in expectation, i.e.
lim E[Xn ] exists. Taking limit as n on both sides of the inequality, we have
n
lim E[Xn ]
2 1
1
1
lim E[Xn1 ] +
lim E[Xn1 ] +
4 n
2 n
4
1
1
1
2
lim E[Xn ] +
lim E[Xn ] +
4 n
2 n
4
2 1
1
1
lim E[Xn ]
lim E[Xn ] +
4 n
2 n
4
2
1
lim E[Xn ] 1
4 n
Hence, lim E[Xn ] = 1. Now we are in the position to prove convergence in probability.
n
Given any > 0,
P (|Xn 1| ) = P (1 Xn )
E[1 Xn ]
1 E[Xn ]
=
where the inequality dues to Markovs inequality. Taking limit on both sides of the
inequality as n , we have
1 E[Xn ]
=0
n
n
lim P (|Xn 1| ) = 0
lim P (|Xn 1| ) lim
p.
Hence, {Xn } 1.
7. Problem 7
a.s., m.s., p., d.
(a) Xn 1
5
v
n n
a.s.
d.
P (V (ny) n1 ), y 0
=
0,
y<0
1
1 e3(ny) n , y 0
=
0,
y<0
Then it can be checked that
FV (y) := lim FVn (y) =
n
1 e3 ,
y0
0,
y<0
which is not a valid cdf ( lim F (y) 6= 1). Hence, {Yn } does not converge in distribution.
y
(c) Zn eV
a.s., p., d.
m.s.
n
p.
Hence, Xn c.
9. Problem 9
(a) Yn has Binomial distribution with parameters (n, n ). Hence, the characteristic function
of Yn , denoted by Yn (u), is
ju n
, uR
Yn (u) = 1 + e
n n
(b) For any u R,
ju n
= exp lim n ln(1 + eju ) = exp((eju 1))
lim Yn (u) = lim 1 + e
n
n
n
n n
n n
which is the characteristic function of Poisson distribution with parameter .
d.
(c) By part (b), Yn Y , where Y has Poisson distribution with parameter . However, we
are lack of information to determine whether Yn converges in other sense or not. For
example, we do not know whether Yn and Ym (m 6= n) are independent or not, since we
did not assume that X1,n and X1,m are independent.