Online Program Home
  My Program

Abstract Details

Activity Number: 353 - Contributed Poster Presentations: Section on Nonparametric Statistics
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
Sponsor: Section on Nonparametric Statistics
Abstract #323464
Title: Conditional Quantiles Under Parametric Model Misspecification: a Simulation Study
Author(s): Bonifride Tuyishimire* and Sergey Tarima and Rodney Sparapani and Lisa Rein and John Meurer
Companies: Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin
Keywords: quantile regression ; model misspecification ; minimum variance ; minimum MSE ; personalized medicine

We compare statistical modelling of conditional quantiles with four different modelling approaches. The first approach uses maximum likelihood estimation under the assumption that the data came from a parametric distribution. The second approach uses the popular nonparametric quantile regression (Koenker, 1978). The third and fourth approaches combine the parametric and nonparametric quantile regressions into a single estimating procedure targeting minimum variance (MVAR) and minimum mean squared error (MMSE) of conditional quantiles, respectively. The MVAR and MMSE show better performance in terms of mean squared errors when data partially follow an assumed parametric model. MVAR outperforms all models at minor departures from the parametric assumptions, whereas MMSE works better when model violations are more severe. We conclude that even partially correct information on parametric family can be effectively used via combined modelling of parametric and non-parametric approaches to improve conditional quantile estimation.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

Copyright © American Statistical Association