|
Activity Number:
|
112
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #304485 |
|
Title:
|
Efficient Estimation for Generalized Linear Model with Varying Dispersion
|
|
Author(s):
|
Zhi He*+ and Douglas Simpson
|
|
Companies:
|
University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign
|
|
Address:
|
Illini 101 725 S.Wright St., Champaign, IL, 61820,
|
|
Keywords:
|
Varying-scale model ; Semiparametric Efficiency ; Heteroscedasticity ; Nonparametric regression ; Linear Exponential Family
|
|
Abstract:
|
This paper deals with the statistical inference based on generalized linear model with nonparametric varying dispersion. When the dispersion and mean of the model are separable, which includes normal, inverse Gaussian and gamma distributions, we present an algorithm to estimate the dispersion nonparametrically, and the asymptotic consistency and normality of the resulting estimator is established. In particular, we show that the asymptotic covariance of the mean parameter can achieve its semiparametric lower bound. This method can be extended to the generalized linear model without separable constraint by using quasi-likelihoods. We illustrated our methodology using both simulated and real data examples.
|
