Bivariate elliptical regression for modeling interval-valued data

This paper introduces a special case of a multivariate regression model with restriction for interval-valued data in the symbolic data analysis framework. This model is less sensitive in the presence of interval outliers since it considers light-heavy tails distributions. Intervals are obtained from...

Full description

Saved in:
Bibliographic Details
Published in:Computational statistics 2022-09, Vol.37 (4), p.2003-2028
Main Authors: Silva, Wagner J. F., Souza, Renata M. C. R., Cysneiros, F. J. A.
Format: Article
Language:English
Subjects:
Bivariate analysis
Data analysis
Datasets
Economic Theory/Quantitative Economics/Mathematical Methods
Intervals
Mathematics and Statistics
Normal distribution
Original Paper
Outliers (statistics)
Probability and Statistics in Computer Science
Probability Theory and Stochastic Processes
Regression models
Statistics
Upper bounds
Variables
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper introduces a special case of a multivariate regression model with restriction for interval-valued data in the symbolic data analysis framework. This model is less sensitive in the presence of interval outliers since it considers light-heavy tails distributions. Intervals are obtained from classic data according to a fusion process and each interval can be represented by its center and range data or lower and upper bound values. The correlation between the center and range variables or lower and upper bound variables is a fundamental component for constructing the model. Therefore, a study that provides a suitable choice of the representation for intervals in bivariate models is proposed. Simulation studies in the Monte Carlo framework regarding different scenarios of interval data set with and without outliers are carried out to validate the proposed model. An application with real-life interval medical dataset is also performed.
ISSN:0943-4062
1613-9658
DOI:10.1007/s00180-021-01189-x