L¨ohr/Winter Winter term 2015/16
Exercises to the lecture Probability Theory II
Exercise sheet 9
Martingale Convergence Theorem
Exercise 9.1 (Moran-model). (4 Points)
LetN, m∈Nwithm < N. We consider a population ofN individuals, out of which initially (in generation 0) m have the (genetic) typea, while the rest has type A. In every timestep (generation) one of the individuals (chosen uniformly at random) dies, and at the same time, a new one is born. Given all previous steps, the probability that the new individual has type a is equal to the relative frequency of typeain the previous generation. Let
Xn := relative frequency ofain the population of thenth generation andT := inf{n∈N|Xn∈ {0,1} }.
(a) Show that (Xn)n∈N0 is a martingale, which converges a.s. and in L1. (b) Show that P {T <∞ }
= 1.
(c) Calculate P {XT = 1} .
Exercise 9.2. (4 Points)
Let (Fn)n∈N be a filtration on a probability space (Ω,A,P). Show that for A ∈ F :=
σ S
n∈NFn
there is a sequence An∈ Fn with
n→∞lim P(An△A) = 0.
Here,An△A= (An\A)∪(A\An).
Hint: Consider Xn:=P(A| Fn) and use the martingale convergence theorem.
Exercise 9.3. (4 Points)
(a) Let a∈R+ and (Xn)n∈N be a martingale withXn+1 ≥Xn−a. Define B :=
lim sup
n→∞
Xn=∞ and C :=
lim inf
n→∞ Xn=−∞ . Show: P(B\C) = 0.
(b) Let Xn,n∈N, be independent withP {Xn=−1}
= n+1n andP {Xn=n}
= n+11 . Show that for Sn:=Pn
k=1Xk: P {lim sup
n→∞
Sn=∞ }
= P {lim inf
n→∞ Sn=−∞ }
= 1.
Hint: Use (a) and Borel-Cantelli.
Please turn
Exercise 9.4(Counter examples). (4 Points) (a) Find a martingale (Mn)n∈N withMn−→
n→∞
∞ a.s.
(b) Find an L1-bounded martingale (Xn)n∈N which is not uniformly integrable.
(c) Letp >1. Find anLp-bounded process (Xn)n∈N, such that |Xn|p
n∈Nis not uniformly integrable.
Due Wed, 13.01. at the beginning of the exercise session
Probability Seminar:
15.12.: Alexander Cox (University of Bath)
Maximising functions of the average of a martingale with given terminal law 22.12.: Sigurd Assing (University of Warwick) .
Tue, 16:15 – 17:15 in WSC-S-U-3.03