# Estimation of Heteroscedasticity in Regression Analysis

The Annals of Statistics
Vol. 15, No. 2 (Jun., 1987), pp. 610-625
Stable URL: http://www.jstor.org/stable/2241329
Page Count: 16

You are not currently logged in.

Access your personal account or get JSTOR access through your library or other institution:

## Abstract

Consider the regression model Yi = g(ti) + εi, 1 ≤ i ≤ n, with nonrandom design variables (ti) and measurements (Yi) for the unknown regression function g(·). We assume that the data are heteroscedastic, i.e., $E(\varepsilon^2_i) = \sigma^2_i \not\equiv \operatorname{const.}$ and investigate how to estimate σ2 i. If σ2 i = σ2(ti) with a smooth function σ2(·), initial estimators σ̃2 i can be improved by kernel smoothers and the resulting class of estimators is shown to be uniformly consistent. These estimates can be used to improve the estimation of the regression function g itself in parametric and nonparametric models. Further applications are suggested.

• 610
• 611
• 612
• 613
• 614
• 615
• 616
• 617
• 618
• 619
• 620
• 621
• 622
• 623
• 624
• 625