Testing the Information Matrix Equality with Robust Estimators



Further,


EFn(Y) = γnφ(0),
E
Fn(Y2)=1+γn + o(γn),
E
Fn(Y 3) = 6γnφ(0) + o(γn),
E
Fn(Y4)=3+6γn+o(γn).

Therefore, letting ( m 1 ,m 2 ,m 3) ' = Wm ( Y ; θn ),
and, upon replacing
γn with ey∕n,

EFn ( m 3)  = EFn

I
--X

sq I

Jλ

×---z

EFn ( m 2)  = EFn

I                                           I

X---X

sq I

Jλ
×---z

EFn ( m^ 1) = EFn

IX )

_

3

_

2

_


6EFn


3EFn


(Y - βn λ

σn J


2

+3=o(γn),


Y—βn = 3 Ynφ (0) + О ( Yn ),


σn


1 - dEFn ( m 3 ) = o ( Yn ),


0

b = lim ∕EFΓ[W [ Wm ( Y ; θn, )] = e   3 φ (0)

n→∞                      0

Hence bV+b = -⅛ e2.

Appendix C

Local asymptotic power of score test

First, we review briefly how the local asymptotic power of the score test
against specified alternatives can be defined. By an appropriate extension
of
f ( ; ), let the density under the alternative be f ( y ; ω ), depending on
an extended parameter
ω, and let s ( y ; ω ) = ∂ω log f ( У ; ω ). Write the
null hypothesis as
H0 : ω Ω0, where Ω0 is a restricted parameter space
(essentially, Θ). Let
ω be the restricted ML estimator (essentially, θ), i.e. ω
solves

n

max    log f (Yi; ω).

ωΩo

i=1

31



More intriguing information

1. Geography, Health, and Demo-Economic Development
2. Fighting windmills? EU industrial interests and global climate negotiations
3. Gerontocracy in Motion? – European Cross-Country Evidence on the Labor Market Consequences of Population Ageing
4. The name is absent
5. The name is absent
6. The name is absent
7. The name is absent
8. Spectral calibration of exponential Lévy Models [1]
9. Change in firm population and spatial variations: The case of Turkey
10. The name is absent