Density Estimation and Combination under Model Ambiguity



1 Σ ln fcn(χl
n
i        g(xi)

Let define


fn (xi) - g(xi)


g(xi)


,  1 X fn(xi) - g(xi) 3 3

+ 2n ¾   g(χi)   J u.


(56)


Vb1n


n ∑(


c(xi) - g(xi)
g(x
i)


and


2
b 1       fn (xi) - g(xi)

v2n = n M   --l x )    ) ■

By Lemma 3.1 Hong-White (2000), under assumption A1 and A2, nh4/ ln n →∞,h0. Then:

KcI1 = Vb1n


13

- V2n+ + Op(n 2h 3lnn + h6).


(57)


AT          1       1         1      JlJ         T^T∙        1 T^T∙    T J 1 1 it X      1 1 —1 T xt—x — —X ZXl       1

Now we have to analyze the terms Vin and V2n. Let define f(x) = ʃh 1K( i- )g(x)dx and

h—iKh(xi
an(xi,xj ) = ----------


- xj ) - fh


iKh (xi - x)g(x)dx


g(xi)


bn(xi) = —


iKhχΛ-χ)f(χ)dx-g(χi)
g(x
i)

Then

Vb1n


c(χi) - f (χi) + f (χi) - g(xi)
g(χi)               g(χi)


n(n ι)h Σ Σ an(xi, xj) + n Σ bn(xi) =
i j,i6=j                       i


(58)


= V11n +B

where Viin is a second order U-statistic and it will affect the asymptotic distribution of KIi. Similarly to

TT ιι<1nnn TJ        ∙. Γ' ■   . ι 1- n ■

Hall(1984) let rewrite Viin in the following way:

Viin = -r
n(n


1y∑ ∑Hin(Xi,Xj)

i j,i6=j


1   Kh (xj

Hin(xi, χj) = 2h (


Xi) - h(x - Xi)g(x)dx + Kh(xi
g(xi)


xj) - Kh(xi - x)g(x)dx

= J                  ] — Jn(xi,xj) + Jn(xj ,xi)

g(xi)

(59)

E(H1n (xi,xj)/xi)=0, then using Theorem 1 in Hall(1984) we can show that

^        I

Viin = < -r
n(n


π∑ ∑Hiin(Xi,Xj)
i j,i6=j

2E [H2n(χi,xj


n2


-)ɪ I →d N(0,1).


(60)


E [J‰1XJ)]   ,h / Kh


xi) - RKh(xi - χ)g(χ)¢2

------g2x )------------------g(xi)g(xj)dxidxj =

26



More intriguing information

1. Towards a Mirror System for the Development of Socially-Mediated Skills
2. The name is absent
3. The name is absent
4. Weather Forecasting for Weather Derivatives
5. Trade Openness and Volatility
6. Monopolistic Pricing in the Banking Industry: a Dynamic Model
7. The name is absent
8. The name is absent
9. Citizenship
10. The name is absent