29.07.2014 Views

Asymptotic Methods in Statistical Inference - Statistics Centre

Asymptotic Methods in Statistical Inference - Statistics Centre

Asymptotic Methods in Statistical Inference - Statistics Centre

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

73<br />

• Example 1. Suppose we (mistakenly) believe that<br />

asample 1 arises from a ( 2 )population.<br />

We test : = 0 vs. : 0 by<br />

us<strong>in</strong>g the fact that if [] =0 2 then (you<br />

should show)<br />

√ ³<br />

<br />

2 − 0<br />

2 ´ ,r<br />

h ( − ) 2i → (0 1)<br />

This does not require normality of the sample. If<br />

the data are normal, then under ,<br />

h ( − ) 2i =2 4 0 = 2 ( 0 ) <br />

so we can take ˆ = √ 20 2 . Suppose that <strong>in</strong> fact<br />

the sample arises from another distribution with<br />

[] =0 2 but h ( − ) 2i = 2 ( 0).<br />

Then<br />

( 0 )<br />

( 0 ) = √ 2<br />

2<br />

0<br />

( 0 ) <br />

which may take on any positive value. Thus any<br />

( ) ∈ (0 12) is atta<strong>in</strong>able and the test is very<br />

non-robust <strong>in</strong> the class F of distributions with f<strong>in</strong>ite<br />

fourth moment (i.e., non-robust aga<strong>in</strong>st nonnormality).

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

Saved successfully!

Ooh no, something went wrong!