Events
Graduate Student Seminar |
| Time: Feb 25, 2015 (03:00 PM) |
| Location: Parker Hall 249 |
|
Details: Speaker: Brice Nguelifack Title: Robust Nonlinear Signed-Rank Regression Abstract: Nonparametric methods have achieved widespread recognition as a valuable technique for analyzing data, particularly data consisting of rank or relative preferences and/or are small samples from unknown distributions. They are also useful for dealing with data that contain aberrant observations since many nonparametric procedures require just the ranks of the observations rather than the actual magnitude of the observations. In this talk, I will give an overview of rank estimation, especially that of regression parameters, and discuss the optimality and efficiency of these estimators. It turns out that the rank method is only slightly less efficient than the best estimator in the normal case and can be more efficient than these competitors when the underlying populations are heavier tailed. Finally, I will discuss rank estimators for signal and image processing models and evaluate efficiencies by way of simulation studies. |