|
Activity Number:
|
354
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #304171 |
|
Title:
|
Penalized Sieve Deconvolution Estimation of Mixture Distributions with Boundary Effects
|
|
Author(s):
|
Mihee Lee*+ and Haipeng Shen and J. Steve Marron
|
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
|
|
Address:
|
Department of Statistics and Operations Research, Chapel Hill, NC, 27599,
|
|
Keywords:
|
Maximum likelihood ; measurement error ; mixture distribution ; sieve method ; penalization
|
|
Abstract:
|
Estimation of the mutation effects distribution is an essential problem in evolutionary biology. However very few statistical approaches have been considered so far. The most common method is a parametric approach based on fitting an exponential distribution whose validity has not been checked. Our approach extends the classical deconvolution setting by allowing the target variable to be a mixture of a point mass and a continuous component. One major contribution of our paper is correct handling of known boundary effects. Moreover, by adopting a roughness penalty, we improve the smoothness of the resulting estimator and reduce the estimation variance. We also propose a graphical tool, the density-envelope plot, to validate the exponential assumption on the mutation effects distribution. We illustrate performances of the proposed estimators via a real application.
|
