SIMEX selection of the smoothing parameter for nonparametric regression with errors-in-variables
Aurore Delaigle (University of Bristol)
15/01/08
In many real life applications, observations can not be measured precisely and the only data available are contaminated by measurement errors (for example, due to the inaccuracy of the measurement device used).
When applied to such data, standard nonparametric estimators of a density or a regression curve are not consistent, but there exist in the literature nonparametric estimators especially adapted to the measurement errors.
These estimators, however, depend crucially on the value of a smoothing parameter. We propose a SIMEX method for selecting this parameter.
The idea can be used in various contexts but, in this talk, we focus on the regression case, where selection of a smoothing parameter in the error context is particularly difficult. This is joint work with Peter Hall.
Back to the seminar homepage
LSE Home Page |
Departmental Home Page |
Baurdoux Home Page
[Last modified: Jan. 4th 2008 by Erik Baurdoux]
