|
Activity Number:
|
291
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #305543 |
|
Title:
|
Efficient Polynomial Spline Estimation of Partially Linear Models for Clustered Data
|
|
Author(s):
|
Lan Xue*+ and Li Wang and Lijian Yang
|
|
Companies:
|
Oregon State University and Oregon State University and Michigan State University
|
|
Address:
|
Kidder 78, Corvallis, OR, 97331,
|
|
Keywords:
|
asymptotic distribution ; knot ; longitudinal data ; milk protein data ; root-n consistency ; semiparametric model
|
|
Abstract:
|
We consider estimation of the partially linear models for clustered data using polynomial spline smoothing. The estimation procedure characterizes the infinitely dimensional nonparametric function by a slowly growing number of parameters. Thus the computation is comparable to parametric least squares. On the other hand, it incorporates the within-cluster correlation properly. The resulting estimators are a "polynomial spline version" of both the profile-kernel (PK) estimators (Lin and Carroll 2001) and backfitting estimators (Zeger and Diggle 1994), replacing kernel smoothing by polynomial spline smoothing. They have the same asymptotic property as the PK estimators. A simulated example demonstrates that the proposed estimators are computationally efficient and as accurate as the PK estimators. Application to milk protein content data is described.
|
