28.01.2015 Views

PDF of Lecture Notes - School of Mathematical Sciences

PDF of Lecture Notes - School of Mathematical Sciences

PDF of Lecture Notes - School of Mathematical Sciences

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2. STATISTICAL INFERENCE<br />

=⇒ f is an exponential family with<br />

A(λ) = −λ<br />

B(λ) = log λ<br />

t(x) = x<br />

h(x) = 0<br />

We can also check that E(X) = −B′ (λ)<br />

A ′ (λ) .<br />

E(X) = 1 λ<br />

for X ∼ Exp(λ).<br />

In particular, we have seen previously<br />

Now observe that A ′ (λ) = −1, B ′ (λ) = 1 λ<br />

=⇒ −B′ (λ)<br />

A ′ (λ)<br />

= 1 , as required.<br />

λ<br />

It also follows that if x 1 , x 2 , . . . , x n are i.i.d. Exp(λ) observations, then ¯X = 1 n<br />

is the MVUE for 1 λ = E(X).<br />

Definition. 2.2.4<br />

n∑<br />

t(x i )<br />

i=1<br />

Consider data with <strong>PDF</strong>/prob. function, f(x; θ). A statistic, S(x), is called a sufficient<br />

statistic for θ if f(x|s; θ) does not depend on θ for all s.<br />

Remarks<br />

(1) We will see that sufficient statistics capture all <strong>of</strong> the information in the<br />

data x that is relevant to θ.<br />

(2) If we consider vector-valued statistics, then this definition admits trivial<br />

examples, such as s = x<br />

⎧<br />

⎨ 1 x = s<br />

since P (X = x|S = s) =<br />

⎩<br />

0 otherwise<br />

which does not depend on θ.<br />

90

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!