Tutorial 4 – MTL390
Problems
1. Let X ∼ Bernoulli(p).
(a) Suppose T (X) is an estimator of ψ(p) = p2 . Write the unbiasedness condition for T .
(b) Show that no unbiased estimator of p2 exists based on a single Bernoulli observation.
2. Let T be an unbiased estimator of θ. Prove that g(T ) is not, in general, an unbiased estimator of
g(θ) unless g is linear.
3. Let X1 , X2 , . . . , Xn be an i.i.d. sample from the uniform distribution on [0, θ]. Show that the
estimator !1/n
n
Y
T (X1 , X2 , . . . , Xn ) = Xi
i=1
is a consistent estimator of θe−1 .
4. Let U be a statistic and let T = g(U ) for some measurable function g. Show that if T is a sufficient
statistic for θ, then U is also sufficient for θ.
5. Let X1 , X2 , . . . , Xn be i.i.d. random variables with common distribution
Pθ,p (x) = (1 − p)p x−θ , x = θ, θ + 1, . . . , 0 < p < 1.
Determine a sufficient statistic in each of the following cases:
(a) p is known and θ is unknown;
(b) θ is known and p is unknown;
(c) both p and θ are unknown.
6. (a) Let X1 , X2 , . . . , Xn be a sample from U [0, θ], θ ∈ Θ = (0, ∞). Let X(n) = max{X1 , X2 , . . . , Xn }.
P
Show that X(n) −
→ θ. Write Yn = 2X. Is Yn consistent for θ?
. , Xn be iid RVs with EXi = µ and E|Xi |2 < ∞. Show that T (X1 , X2 , . . . , Xn ) =
(b) Let X1 , X2 , . . P
−1 n
2[n(n + 1)] i=1 iXi is a consistent estimator for µ.
7. Find a sufficient statistic in each of the following cases based on a random sample of size n:
(a) X ∼ B(α, β) when (i) α is unknown, β known; (ii) β is unknown, α known; and (iii) α, β are
both unknown.
(b) X ∼ G(α, β) when (i) α is unknown, β known; (ii) β is unknown, α known; and (iii) α, β are
both unknown.
8. Let T1 , T2 be two statistics with common domain D. Then show that T1 is a function of T2 if and
only if
for all x, y ∈ D, T1 (x) = T1 (y) =⇒ T2 (x) = T2 (y).
9. Let X1 , X2 , . . . , Xn be a random sample of size n from a Normal distribution N (µ, σ02 ), where the
mean µ is the unknown parameter of interest and the variance σ02 is known.
Use the Neyman-Fisher Factorisation Criterion to show that the sample mean T (X1 , . . . , Xn ) = X
is a sufficient statistic for µ.
1
10. Let X1 , X2 , . . . , Xn be a random sample from a Poisson distribution with parameter λ. Show that
αX̄ + (1 − α)s2 , 0 ≤ α ≤ 1,
is a class of unbiased estimators for λ. Also, find an unbiased estimator for e−λ .
11. Let X1 , X2 , . . . , Xn be a random sample from the normal distribution N (µ, σ 2 ).
(a) Suppose that σ 2 is known. Find the minimal sufficient statistic for the parameter µ.
(b) Suppose that µ is known. Find the minimal sufficient statistic for the parameter σ 2 .
Pn
12. Let X1 , . . . , Xn be a random sample from a distribution with parameter θ. Let T (X) = i=1 Xi be
a known sufficient statistic for θ.
Consider the transformed statistic T ∗ = (T (X))2 . Determine whether T ∗ is sufficient for θ in the
following two cases:
(a) Xi ∼ N (θ, 1) where θ ∈ R.
(b) Xi ∼ Exp(θ)
(Intuitively, think about sufficiency of a statistic as ”containing all the information about the pa-
rameter”)
13. Let X1 , X2 , . . . , Xn be a random sample from a distribution with the probability density function
(
θxθ−1 , 0 < x < 1,
f (x; θ) =
0, otherwise,
where θ > 0. Find a sufficient statistic for θ.
14. Let X1 , . . . , Xn ∼ N (µ, σ 2 ). By the Factorization Theorem, we know that the vector of raw sums:
n n
!
X X
2
T = (T1 , T2 ) = Xi , Xi
i=1 i=1
is jointly sufficient for the parameter vector (µ, σ 2 ).
Show that the pair T ′ = (X̄, S 2 ) is also sufficient, where S 2 is the sample variance defined as:
n
1 X
S2 = (Xi − X̄)2
n−1
i=1