Scribd
Upload
10 views2 pages

Exercise Sheet 2

The document is an exercise sheet for a Mathematical Statistics course, detailing various statistical concepts and problems related to sufficiency, method of moments, and maximum-likelihood estimators. It includes exercises on the Neyman criterion for sufficiency, proving sufficiency with measurable mappings, determining point estimators, and factorization of maximum-likelihood estimators. The exercises involve log-normal and Pareto distributions, as well as specific density functions and their implications for statistical modeling.

Uploaded by

ola kuke
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
10 views2 pages

Exercise Sheet 2

The document is an exercise sheet for a Mathematical Statistics course, detailing various statistical concepts and problems related to sufficiency, method of moments, and maximum-likelihood estimators. It includes exercises on the Neyman criterion for sufficiency, proving sufficiency with measurable mappings, determining point estimators, and factorization of maximum-likelihood estimators. The exercises involve log-normal and Pareto distributions, as well as specific density functions and their implications for statistical modeling.

Uploaded by

ola kuke
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/ 2

PD Dr. Lorenz A.

Gilch
Chair of Stochastics
und its Applications

Exercises Mathematical Statistics“


”
Summer semester 2022
Exercise sheet 2
19.5.2022

1. Neyman criterion for sufficiency:


Let be Pθ = LN(µ, σ 2 ) with parameter θ = (µ, σ 2 ) ∈ R × (0, ∞) , that is, Pθ is the
log-normal distribution on (R, B) with Lebesgue density
2 !
(1) 1 ln(x) − µ
fθ (x) = √ x−1 exp − 1(0,∞) (x), x ∈ R.
2πσ 2 2σ 2

Fix now any n ∈ N. Let X1 , . . . , Xn be independent and identically distributed random


variables with X1 ∼ Pθ . Then X := (X1 , . . . , Xn ) is a random vector with image X = Rn .
Show that
X n n 
2
X 2
T : X → R , (x1 , . . . , xn ) 7→ ln(xi ) , ln(xi ) .
i=1 i=1

is a sufficient statistic for θ in the statistical model X , P , where
 n 
(1)
Y
n
P= fθ : R → [0, ∞) ∀x1 , . . . , xn ∈ R : fθ (x1 , . . . , xn ) = fθ (xi ) .
i=1

2. Sufficiency:
Prove Exercise 2.9 from the lecture:
Let X be a random variable with (discrete or continuous) density fθ in dependence of some
parameter θ. Furthermore, let T : X → T be a sufficient statistic for θ and let b : T → T be
a bijective, measureable mapping which is independent from the parameter θ. We assume
also that b−1 is measurable.
Then b ◦ T also is a sufficient statistic for θ.

3. Method of moments:
Consider the Lebesgue density
2x
fθ (x) = 1[0,θ] (x) · , x ∈ R.
θ2

Let X1 , . . . , Xn be i.i.d. random variables with density fθ . Determine a point estimator θ̂


for θ via the method of moments.

Please turn −→
4. Maximum-likelihood estimator:

a) Let X1 , . . . , Xn ∼ Pa(1, θ) be i.i.d. random variables, where Pa(1, θ) is the Pareto


distribution with parameters 1 and θ ∈ (0, ∞). Determine the ML estimator for θ.
Recall:
The Lebesgue density of Pa(c, θ) is given by fc,θ (x) = θcθ x−(θ+1) 1(c,∞) (x) with para-
meters c, θ, > 0.

b) Let X1 , X2 , . . . , Xn be i.i.d. real-valued random variables. Set α := exp − exp(exp(100)) .
Assume that the distribution of X1 has the Lebesgue density
1−α  1  α  1 
fθ (x)(1) = √ exp − (x − a)2 + √ exp − 2 (x − a)2
2π 2 2πσ 2 2σ

where θ := (a, σ 2 ) ∈ R × (0, ∞) =: Θ. By definition of α, Xi is approximately


N (a, 1)-distributed. Show that
n
Y
n
∀(x1 , . . . , xn ) ∈ R : sup fθ (xi ) = ∞,
θ∈Θ i=1

which implies that the ML estimator for θ does not exist in the model (Rn , P), where
(1)
P = fθ : Rn → [0, ∞) | ∀x1 , . . . , xn ∈ R : fθ (x1 , . . . , xn ) = ni=1 fθ (xi )
 Q

Hint: Choose a = x1 and let σ 2 & 0.

5. Factorisation of ML estimator:
Let X : Ω → X be a random variable with (discrete/continuous) density fθ and let
T : X → T be a sufficient statistic for the parameter θ ∈ Θ. Show that the maximum-
likelihood estimator θ̂M L factorizes over T , that is, there exists a function g such that
θ̂M L = g ◦ T .

You might also like