Efficient Bandwidth Selection in Non‐parametric Regression (Q4455936)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 2059374
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Efficient Bandwidth Selection in Non‐parametric Regression |
scientific article; zbMATH DE number 2059374 |
Statements
Efficient Bandwidth Selection in Non‐parametric Regression (English)
0 references
16 March 2004
0 references
local linear regression
0 references
asymptotic mean squared error
0 references
asymptotic normality
0 references
0.8737303018569946
0 references
0.8586292862892151
0 references
0.8586125373840332
0 references
0.8584581017494202
0 references
0.8535923361778259
0 references
For a bivariate sample \((X_i,Y_i)\), \(i=1,\dots,n\), a regression function \(m(x)=E(Y \,|\,X=x)\) is estimated by the local linear estimator \[ \widehat m_k(x,b)=\arg \min_{\beta_0} \sum_{i=1}^n(Y_i-\beta_0-\beta_1(x-X_i))^2K((x-X_i)/b), \] where \(K\) is a kernel of order \(k\), and \(b\) is a bandwidth. Asymptotics of \(\widehat m_k\) and its asymptotic MSE are investigated. A two-stage procedure of optimal (adaptive) bandwidth selection is proposed where the main part of the estimated AMSE is minimized at each step. For the estimation of the bias in the AMSE estimator the difference \(\widehat m_2-\widehat m_4\) is used. The asymptotic normality of the normalized \(\widehat m_2\) with adaptive bandwidth is demonstrated. Results of simulations are presented.
0 references