Part A Probability, Michaelmas Term 2024 James Martin

Problem Sheet 1

1. A company sells lottery scratch-cards for £1 each. 1% of cards win the grand prize of
£50, a further 20% win a small prize of £2, and the rest win no prize at all. Estimate
how many cards the company needs to sell to be 99% sure of making an overall profit.
[Φ(2.3263) = 0.99]

2. A list consists of 1000 non-negative numbers. The sum of the entries is 9000 and the sum
of the squares of the entries of 91000. Let X represent an entry picked at random from
the list. Find the mean of X, the mean of X 2 , and the variance of X. Using Markov’s
inequality, show that the number of entries in the list greater than or equal to 50 is at
most 180. What is the corresponding bound from applying Markov’s inequality to the
random variable X 2 ? What is the corresponding bound using Chebyshev’s inequality?

3. Let Yn be uniform on {1, 2, . . . , n} (i.e. taking each value with probability 1/n). Draw the
distribution function of Yn /n. Show that the sequence Yn /n converges in distribution as
n → ∞. What is the limit?

4. Let Xi , i ≥ 1 be i.i.d. uniform on [0, 1]. Let Mn = max{X1 , . . . , Xn }.

(a) Show that Mn → 1 in probability as n → ∞.

5. (a) What is the distribution of the sum of n independent Poisson random variables each
of mean 1? Use the central limit theorem to deduce that
n2 nn
 
1
e−n 1 + n + + ··· + → as n → ∞.
2! n! 2

(b) Let p ∈ (0, 1). What is the distribution of the sum of n independent Bernoulli random
variables with parameter p? Let 0 ≤ a < b ≤ 1. Use appropriate limit theorems to
determine how the value of
bbnc
X n
lim pr (1 − p)n−r
n→∞ r
r=banc

depends on a and b. (bxc is the integer part of x.)

6. (a) Let Xn , n ≥ 1 be a sequence of random variables defined on the same probability

7. A gambler makes a long sequence of bets against a rich friend. The gambler has initial
capital C. On each round, a coin is tossed; if the coin comes up tails, he loses 30% of his
current capital, but if the coin comes up heads, he instead wins 35% of his current capital.

(a) Let Cn be the gambler’s capital after n rounds. Write Cn as a product CY1 Y2 . . . Yn
where Yi are i.i.d. random variables. Find E Cn .
(b) Find the median of the distribution of C10 and compare it to E C10 .
(c) Consider log Cn . What does the law of large numbers tell us about the behaviour of
Cn as n → ∞? How is this consistent with the behaviour of E Cn ?

8. Let Hn be the n-dimensional cube [−1, 1]n . For fixed x ∈ R, show that the proportion
of the volume of Hn within distance (n/3)1/2 + x of the origin converges as n → ∞,
and find the limit. [Hint: Consider a random point whose n coordinates are i.i.d. with
Uniform[−1, 1] distribution. If A ⊂ Hn , then vol(A)/vol(Hn ) is the probability that such
a point falls in the set A. Let Dn represent the distance of such a point from the origin;
apply an appropriate limit theorem to Dn2 .]

9. Let Y1 , Y2 , . . . be i.i.d. and uniformly distributed on the set {1, 2, . . . , n}. Define X (n) =
min{k : Yk = Yj for some j < k}, the first time that we see a repetition in the sequence
√
Yi . (Interpret the case n = 365). Prove that X (n) / n converges in distribution to a limit
with distribution function F (x) = 1 − exp(−x2 /2) for x > 0.
[Hint: Observe that
   1

2
 
m−1

(n)
P X >m = 1− 1− ... 1 − .
n n n

You may find it useful to use bounds such as −h − h2 < log(1 − h) < −h for sufficiently
small positive h. ]

Additional problems:

10. Let Xi , i ≥ 1 be i.i.d. random variables with P(Xi = 0) = P(Xi = 1) = 1/2.

(a) Define Sn = ni=1 Xi 2−i . What is the distribution of Sn ? Show that the sequence
P
Sn converges almost surely as n → ∞ (hint: Cauchy sequences converge). What is
the distribution of the limit?
Pn −i
(b) Define Rn = i=1 2Xi 3 . Show again that the sequence Rn converges almost
surely. Is the limit a discrete random variable? Is it a continuous random variable?
(Consider its expansion in base 3).

11. Let An be the median of 2n + 1 i.i.d. random variables which are uniform on [0, 1]. Find
the probability density function of An . [Hint: consider the probability
 that
 An lies in
1 a
a small interval (x, x + dx).] How does the density at the point +√ behave as
2 n
n → ∞? (Stirling’s formula may be useful). Deduce a convergence in distribution result
for the median (appropriately rescaled) as n → ∞ (feel free to argue informally if you
like!).