0% found this document useful (0 votes)
4 views4 pages

2021 Sol

KU Leuven Probability and Measure, final exam

Uploaded by

voteveto1
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
0% found this document useful (0 votes)
4 views4 pages

2021 Sol

KU Leuven Probability and Measure, final exam

Uploaded by

voteveto1
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/ 4

KU LEUVEN

DEPARTEMENT WISKUNDE
Exam Name:
G0P63B Probability and Measure
17 January 2021 (200C 01.17, 12:00–15:00)
Answer the questions below. Only your best three out of four solutions count
towards your grade. Write your name on the top of every sheet you submit.
This is an open book exam: You are allowed to use the official lecture notes
and the exercise sheets without solutions. Marks may be deducted for written
answers that do not clearly show how the solution is reached. Use complete
grammatical sentences in the English language.
Please place your student ID card on the table at a distance from yourself for
inspection, with the picture facing up. If you need more sheets of paper, raise
your hand to request it. GOOD LUCK!!!

1. Let f : (0, 1) ! C be a Lebesgue-integrable function. Consider the function


Z 1
f (y) sin(4xy)y
' : R ! C, '(x) = d (y).
0 1 + x2 + y 3
(a) Argue why ' is a well-defined map.
(b) Is ' continuous? Justify your answer!
(c) Is ' integrable? Justify your answer!

2. Let (⌦, M, µ) be a measure space and fn : ⌦ ! C a sequence


R of measurable
functions. Suppose that p > 0 is a number such that ⌦ |fn | dµ  n1p for all
n 1. When can one conclude that limn!1 fn (x) = 0 for µ-almost all x 2 ⌦?
Justify your answer in detail, and keep it mind that it may depend on p.

3. Let X and Xn , n 1, be real random variables defined over the probability


space (⌦, M, P).
(a) Suppose Xn converges to X in probability. Show that Xn converges to X
in distribution.
Hint: It is useful to formulate the claim in terms of distribution functions.
(b) Suppose that Xn converges to zero in distribution. Show that Xn also
converges to zero in probability.

4. Let (⌦, M, P) be a probability space and An 2 M an independent Pnsequence of


1
events. For all n 1, consider the
P real random variable Y n = n k=1 Ak and
the probability average pn = n1 nk=1 P(Ak ). Show that (Yn pn ) converges to
zero in probability.
Is this a special case of the weak law of large numbers? If not, then what
assumption could one add to make it a special case? Justify your answer!
KU LEUVEN
DEPARTEMENT WISKUNDE
Exam Name: SOLUTIONS
G0P63B Probability and Measure
17 January 2021 (200C 01.17, 12:00–15:00)
1. (a) ' is well-defined because the function fx := [y 7! f (y) sin(4xy)y
1+x2 +y 3
] is again Lebesgue-
integrable. We have |fx |  |f | and therefore
Z 1 Z 1
f (y) sin(4xy)y
d (y)  |f | d < 1,
0 1 + x2 + y 3 0

so the integral used to define '(x) really exists for all x 2 R.


(b) Yes, ' is continuous. This is a consequence of the Dominated Convergence
Theorem. We have |fx |  |f | for all x 2 R and fx (y) ! fx0 (y) as x ! x0 for all
y 2 R. If xn 2 R is any sequence converging to x 2 R, then it follows that
Z 1 Z 1
f (y) sin(4xn y)y f (y) sin(4xy)y f (y) sin(4xn y)y f (y) sin(4xy)y
2 + y3 2 + y3
d (y)  d (y) ! 0.
0 1 + x n 1 + x 0 1 + x2n + y 3 1 + x2 + y 3
| {z }
!0

R
(c)
R Yes, ' is integrable. Indeed, it follows from the general integral formula | (...)| 
|...| and Fubini’s theorem that ' is integrable if one has
Z 1⇣Z ⌘
f (y) sin(4xy)y
2 3
dx dy < 1.
0 R 1+x +y
We have
f (y) sin(4xy)y |f (y)y| 1
2 3
 3
· x2
.
1+x +y 1+y 1 + 1+y 3

1
Since the function 1+x 2 is well-known to be integrable over R, let us say that C > 0
is its integral. Then
R f (y) sin(4xy)y |f (y)y| R 1
R 1+x2 +y 3
dx  1+y 3
· R 1+ x2
dx
1+y 3
|f (y)y|
= 1+y 3
(1 + y 3 )1/2 C
|f (y)y|
= C · p 3  C|f (y)|.
1+y

Since we assumed f to be integrable, the claim follows.

P1 1
2. Below we will use the standard fact from first year calculus that n=1 np < 1 if
and only if p > 1.
1
If p > 1, then the answer is yes. Notice
P1 ⇣ R ⌘ assumption, we have fn 2 L (µ)
that by
for all n 1, and the series n=1 ⌦ |fn | dµ is absolutely convergent. By the
R P1
Monotone Convergence Theorem,
P1 we thus have ⌦ n=1 |fn | dµ < 1. By Propo-
sition 1.38, it follows that n=1 |fn (x)| converges for µ-almost all x 2 ⌦. For such
points, it of course holds that fn (x) ! 0, which finishes the claim.
G0P63B Probability and Measure Exam Name: 27.1.021 page 2

On the other hand, if p  1, then the conclusion would be false in general. It suffices
to disprove the case p = 1. A counterexample is given by ⌦ = [0, 1) being equipped
with the restriction of the Lebesgue measure. Consider the sequence of functions fn
given by fn = 12 [k2 ` ,(k+1)2 ` ) , where k, l 0 are the unique numbers such that k <
R
2` and n = 2` + k. Then we can see ⌦ |fn | d = 12 ([k2 ` , (k + 1)2 ` )) = 12 2 ` < n1 .
However, the sequence fn converges nowhere. Indeed, we see that fn and fn+1 have
disjoint support for all n 2, and for every x 2 [0, 1), there are infinitely n such
that fn (x) 6= 0. This implies that fn (x) cannot converge.

3. (a) In order to show that Xn converges to X in distribution, we need to show by


Theorem 3.37 that for the distribution functions Fn of Xn and F of X, we have
Fn (t) ! F (t) whenever t 2 R is a continuous point of F .
Let " > 0 and t 2 R. Then we observe for every n 1 that

Fn (t) = P(Xn  t)  P(X  t + ") + P(|Xn X| > ") = F (t + ") + P(|Xn X| > ")

and

F (t ") = P(X  t ")  P(Xn  t) + P(|Xn X| > ") = Fn (t) + P(|Xn X| > ").

In summary, we have the inequality

F (t ") P(|Xn X| > ")  Fn (t)  F (t + ") + P(|Xn X| > ").

Since we assumed that Xn converges to X in probability, the probabilities appearing


here go to zero as n ! 1. If t is a point of continuity of F , then lim"!0 F (t ± ") =
F (t), hence this shows that indeed limn!1 Fn (t) = F (t).
(b) We assume Xn converges to zero in distribution. Note that the measure as-
sociatedR to the zero variable is the Dirac measure 0 on R. By ExS-11-3, this
means R f dPXn ! f (0) for every bounded continuous function f : R ! C.
Given any " > 0, choose any bounded continuous non-negative function f" such
that f" R\[ ","] and f" (0) = 0. Then observe observe that by definition of the
measures PXn , we have
Z Z
P(|Xn | > ") = PXn (R \ [ ", "]) = R\[ ","] dPXn  f" dPXn ! 0.
R R

Since " > 0 was arbitrary, this proves that Xn converges to zero in probability.

4. Before we get into the details, let us already justify one fact: The events An 2 M
form an independent sequence if and only if An form an independent sequence of
real random variables. The “if” part is clear, and the “only if” part follows from the
fact that the -algebra associated to the real random variable An is equal to all of
its preimages, so {;, An , ⌦ \ An , ⌦}. If the An form an independent sequence, then it
follows by ExS-9-4 that the sequence of these -algebras is also independent, which
by definition means that the An form an independent sequence of real random
variables.
We can already answer the second part of the problem. In general, the claim is not
a special case of the weak law of large numbers, because the real random variable
An has mean P(An ), which does not have to be equal for all n 1. On the other
hand, if we do assume that P(An ) = p 2 R for all n 1, then the real random
variables An become identically distributed, and one has pn = m for all n 1. So
in this case, the weak law of large numbers indeed implies that Yn p converges to
zero in probability.
Now let us prove the general case. Note that we have E(Yn ) = pn for all n 1.
Not unlike how we proved the weak law of large numbers, we start computing the
variance of the involved real random variables. We have

1 X ⇣ ⌘
n
Var(Yn ) = 2 E ( Aj P(Aj ))( Ak P(Ak ))
n
j,k=1

The real random variables Aj P(Aj ) are mutually independent for all j 1, so it
follows from Proposition 3.25 that the summand over the pair (j, k) vanishes when
j 6= k. In particular,

1 X ⇣ ⌘
n
1
Var(Yn ) = E ( Aj P(Aj ))2  .
n2 | {z } n
j=1
1

Using Corollary 3.27(ii), we have for all " > 0 that


1 1 n!1
P(|Yn pn | ")  2
Var(Yn )  2 ! 0.
" n"
Since " > 0 was arbitrary, this fi
nishes the proof.

G0P63B Probability and Measure 17 January 2021

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy