Abstract Details
Activity Number:
|
194
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 5, 2013 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Biometrics Section
|
Abstract - #309363 |
Title:
|
Local Likelihood-Based Estimation for Quantile Classification in the Logistic Regression Model
|
Author(s):
|
John Rice*+ and Jeremy Taylor
|
Companies:
|
University of Michigan and University of Michigan
|
Keywords:
|
classification ;
robust estimation ;
logistic regression ;
asymmetric loss ;
local likelihood
|
Abstract:
|
Much work has been done on the problem of predicting future binary outcomes in a population based on a sample from that population. However, most authors have focused primarily on median classification (i.e., a positive response is predicted if $\hat{p}_i>0.5$) or have made use of nonparametric ``black box'' methods. We desire to classify future subjects on the basis of a probability threshold $p^*$, not necessarily equal to 0.5, using a rule based on a linear combination of the covariates. To do this, we solve a weighted form of the score equations, using a kernel-like weight function centered about $p^*$; the bandwidth for the weight function is selected by cross validation. This work differs from most previous approaches in local likelihood and robust methods in that the weight depends on both the unknown regression parameters and the covariates (but not the outcome). Simulation results are given showing the reduction in misclassification rates that can be obtained with this method, even under certain forms of model misspecification. Analysis of a melanoma data set is also presented to demonstrate the utility of the method in practice.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2013 program
|
2013 JSM Online Program Home
For information, contact jsm@amstat.org or phone (888) 231-3473.
If you have questions about the Continuing Education program, please contact the Education Department.
The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.
Copyright © American Statistical Association.