Linear regression estimation in non-linear single index models

Authors Fadoua Balabdaoui, Gian-Andrea Thanei

Publication date 22 December 2016

Abstract: In this article, we consider the problem of estimating the index parameter

a l p h a_{0}

in the single index model

E [Y | X] = f_{0} (a l p h a_{0}^{T} X)

with

f_{0}

the unknown ridge function defined on

m a t h b b R

,

X

a d-dimensional covariate and

Y

the response. We show that when

X

is Gaussian, then

a l p h a_{0}

can be consistently estimated by regressing the observed responses

Y_{i}

,

i = 1, ..., n

on the covariates

X_{1}, ..., X_{n}

after centering and rescaling. The method works without any additional smoothness assumptions on

f_{0}

and only requires that

c o v (f_{0} (a l p h a_{0}^{T} X), a l p h a_{0}^{T} X) e q 0

, which is always satisfied by monotone and non-constant functions

f_{0}

. We show that our estimator is asymptotically normal and give the expression with its asymptotic variance. The approach is illustrated through a simulation study.

This page was built for publication: Linear regression estimation in non-linear single index models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6281168)