Scribd
Upload
5 views

Ferguson3

The document presents solutions to Exercise Set 2 from Stat 200C, detailing probability calculations and characteristic functions. It discusses the least squares estimates and their consistency in quadratic mean, providing mathematical derivations and conditions for various estimates. The results highlight the relationships between different estimators based on the behavior of the weights used in the calculations.

Uploaded by

Sambsamb Sambiani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
5 views

Ferguson3

The document presents solutions to Exercise Set 2 from Stat 200C, detailing probability calculations and characteristic functions. It discusses the least squares estimates and their consistency in quadratic mean, providing mathematical derivations and conditions for various estimates. The results highlight the relationships between different estimators based on the behavior of the weights used in the calculations.

Uploaded by

Sambsamb Sambiani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Stat 200C, Spring 2010 Ferguson

Solutions to Exercise Set 2.


2.7. P(N = n) = P(Xn = 1, Xn+1 = 0, Xn+2 = 0, . . .)
= P(Xn = 1)P(Xn+1 = 0)P(Xn+2 = 0) · · ·
1 n(n + 2) (n + 1)(n + 3) 1
= 2· 2
· 2
··· =
n (n + 1) (n + 2) n(n + 1)
∞ ∞
So E(N ) = 1 n/(n(n + 1)) = 1 1/(n + 1) = ∞.

3.5. (a) The characteristic function is


 ∞  ∞
1 1
α
xα−1 itx−(x/β)
e dx = α
xα−1 e−((1−itβ)/β))x dx
Γ(α)β 0 Γ(α)β 0
 α
1 β 1
= Γ(α) =
Γ(α)β α 1 − itβ (1 − itβ)α
(b) The characteristic function of Xn − n is

ϕXn−n (t) = e−itn (1 − (it/n))−n .


2

It is sufficient to show that log ϕXn −n (t) → −t2 /2.

log ϕXn −n (t) = −itn − n2 log(1 − (it/n))


= −itn + n2 [(it/n) + (1/2)(it/n)2 + O(n−3 )]
= −t2 /2 + O(n−1 ) → −t2 /2
n
4.1. (a) The least squares estimate of λ is that value of λ that minimizes 1 (Xi −λzi )2 .
Taking
n a derivative
n 2 of this sum with respect to λ, setting to zero and solving gives λ̂LS =
X z / z . The weighted least squares estimate of λ is that value of λ that minimizes
1n i i 1 i n n
1 (Xi − λzi ) /zi . Similarly, we find λ̂W =
2
1 Xi / 1 zi .
(b) Since Eλ̂LS = λ, it is sufficient to find for what values of the zi we have Var(λ̂LS ) →
0 as n → ∞ (see Exercise 1.5). But
n n n 2 n
Xi zi Var 1 Xi zi zi VarXi λ 1 zi3
Varλ̂LS = Var n 2 =
1 n = n
1
= n 2 2
1 zi ( 1 zi2 )2 ( 1 zi2 )2 ( 1 zi )
n 3 n 2 2
Therefore, λLS is consistent in quadratic mean if and only if 1 zi /( 1 zi ) → 0 as
n → ∞. n n n
(c) λ̂W is also unbiased, and Var(λ̂W )= λ 1 zi /( 1 zi )2 = λ/ 1 zi . So λ̂W is
n
consistent in quadratic mean if and only if 1 zi → ∞ as n → ∞.
∞ ∞
(d) If λ̂W is not consistent in quadratic mean, i.e. if 1 zi < ∞, then both 1 zi2 <
∞ ∞
∞ and 1 zi3 < ∞, so λ̂LS is not consistent either. However if zi = 1/i, then 1 zi = ∞
∞ ∞
so that λ̂W is consistent in quadratic mean, but 1 zi2 < ∞ and 1 zi3 < ∞ so that λ̂LS
is not consistent.

You might also like