Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 57 - Nonparametric Modeling I
Type: Contributed
Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
Sponsor: Section on Nonparametric Statistics
Abstract #319214
Title: A Robust Non Parametric Median-Based Quantile Regression Control Chart Using Real-Life Case Studies
Author(s): Quratulain Khaliq*
Companies: Bank of Punjab
Keywords: Non-Parametric Statistics; Quantile Regression; Regression Control Chart; Outlier Detection Methods
Abstract:

A regression control chart is a powerful tool from the SPC toolkit. However, in the case of an influential outlier, the response variable is affected by an explanatory variable; this design no longer is reliable when homogeneity assumption is violated; therefore, need of such designs which are robust to an outlier or best for heterogeneity effects of covariates through conditional quantities of the outcomes variables and provide abroad image of the whole distribution of response variables. Quantile regression (QR) has gained increasing attention in recent years, and it is becoming more applicable in a variety of fields, including finance, ecology, healthcare, economics, financial-economics, industries, engineering, statistics, medicine, and growth charts, where it is used to screen the percentile curves of abnormal growth. The majority of practitioners considered it because of its effectiveness in dealing with data instabilities or extremes. This study has provided a comparative performance study of regression versus Quantile regression using three case studies; case study one is about the relationship of hardness in Rockwell and tensile strength; case study two is about simulation study by introducing influential outliers in data, the third case study is from fuel consumption data of cars. This study further provided a brief performance comparison of regression versus quantile regression. Quantile regression is robust in all cases, and its residual performance was far better than regression. Even quantile regression charts are also more sensitive and have better performance than regression charts.


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

Back to the full JSM 2021 program