Mixed-type distribution plots. (July 2019)

• Record Type: Journal Article
• Authors: Weld, Christopher; Leemis, Lawrence
• Is Part Of: Information visualization. Volume 18:Number 3(2019)
• Page Start: 311
• Page End: 317
• Publication Date: 2019-07
• Subjects: Algorithms -- data analysis -- data visualization -- graphical perception -- multidimensional scaling -- plots -- statistical graphics -- visual data analysis; Information visualization -- Periodicals; 006.605
• DOI: 10.1177/1473871618756584
• Languages: English
• ISSNs: 1473-8716

Abstract

Plotting is among the most effective ways to quickly and accurately describe a probability distribution. It makes often complex information accessible, enabling intuition for respective outcomes at a glance. Matters complicate, however, for mixed-type distributions. Mixed-type distributions contain both continuous and discrete components, and accurately portraying those on a single axis can prove difficult—misleading intuition as a consequence of pulling two otherwise disjoint components into focus together. This article examines the challenges of maintaining the simple, concise, and accurate format of traditional probability distribution plots for mixed-type distributions. We illustrate issues arising within this plot classification paradigm, and why a secondary axis is uniquely suited to improve its communication. An algorithm is devised to consistently scale such plots so that they better coincide with intuition. National Football League football starting field position, meteorological data, and financial instruments provide examples demonstrating effectiveness of this plot technique.