Evaluating the Effects of Misspecification in the Symbolic Linear Regression for Interval-Valued Data

• Activity Number: 422 - Contributed Poster Presentations: Social Statistics Section
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Natalia Costa Araujo* and Lynne Billard
• Companies: University of Georgia and University of Georgia
• Keywords: Symbolic data; Interval-valued data; Linear Regression; Misspecification

Abstract

Symbolic data analysis was first introduced by Diday (1987) and presents an alternative approach to classical data when the data have a more complex formulation. Even though symbolic data happen more commonly due to the aggregation of large datasets, it can also happen naturally. It can assume various forms, such as interval-valued data, histogram-valued data, lists, among other complex formats. There is an increasing need to develop and improve techniques to deal with and make inferences about symbolic data, while offering efficiency and interpretability to the results. Focusing solely on interval-valued data, this project intends to understand the behavior of the estimation of linear regression methods developed under the assumption of uniformity within the interval, when this assumption is not met. Simulation studies have been conducted to comprehend the extent of the effect of the uniformity assumption, and how the estimates perform when the data within the interval is normally distributed. These results are compared for different scenarios, while modifying the sample size and the width of the interval.