Guidelines for Presenting Quantitative Data in HFES Publications

Abstract

This article provides guidelines for presenting quantitative data in papers for publication. The article begins with a reader-centered design philosophy that distills the maxim “know thy user” into three components: (a) know your users′ tasks, (b) know the operations supported by your displays, and (c) match user's operations to the ones supported by your display. Next, factors affecting the decision to present data in text, tables, or graphs are described: the amount of data, the readers′ informational needs, and the value of visualizing the data. The remainder of the article outlines the design decisions required once an author has selected graphs as the data presentation medium. Decisions about the type of graph depend on the readers′ experience and informational needs as well as characteristics of the independent (predictor) variables and the dependent (criterion) variable. Finally, specific guidelines for the design of graphs are presented. The guidelines were derived from empirical studies, analyses of graph readers′ tasks, and practice-based design guidelines. The guidelines focus on matching the specific sensory, perceptual, and cognitive operations required to read a graph to the operations that the graph supports.

Effects of spatial compatibility on integration processes in graph comprehension

“…The concept of similarity, which was already introduced as a Gestalt principle in the seminal work of Max Wertheimer (1923;see also Ware, 2004), can refer to perceptual attributes such as color and texture, but also to spatial position (see Wickens & Hollands, 2000). More specifically, Gillan et al (1998) explicitly predicted that the ordering of legend symbols should match the order of indicators in the data region, a claim that is directly supported by our data.…”

“…Previous guidelines have recommended bar graphs for depicting differences between means across different levels of the independent variable and when the latter represents an ordinal or categorical variable (Gillan et al, 1998). However, our between-experiment comparison suggests a disadvantage for bar graphs even for the extraction of main effects in basic 2 × 2 designs.…”

“…Thus, for effective integration processes, any difficulties with regard to legend or title should be minimized-for example, by avoiding legends through direct labeling of bars and lines (Gillan, Wickens, Hollands, & Carswell, 1998;Kosslyn, 1994;Schmid, 1983). However, in complex or dense graphs, direct labeling easily results in visual cluttering of textual elements and, thus, cannot be considered a panacea.…”

Thinking Graphically: Connecting Vision and Cognition during Graph Comprehension

“…Several papers provide recommendations for improving graph design (Bertin, 1983;Carpenter & Shah, 1998;Gillian et al, 1998;Kosslyn, 1989;Shah & Carpenter, 1995;Shah, Mayer, & Hegarty, 1999;Shah & Hoeffner, 1992;Pinker, 1990). Our framework can also be used to provide specific guidelines for improving integration processes in graphs; specifically, we focus on recommendations for complex visualizations (primarily color-coded graphs) where several data points need to be represented.…”

Usability Science. I: Foundations

“…Lohse's (1993) studies and model served to guide the development of a computer-based tool that analyzes graphs and makes recommendations for improvements based on performance predicted by the model. The research described previously by Gillan, Carswell, and Wickens (Gillan, Wickens, Hollands, & Carswell, 1998) served as an important basis for a set of guidelines for graph design for publications of the Human Factors and Ergonomics Society. Finally, Gillan and Callahan (2000) used their model and research on pie graphs to develop new versions of the pie graph designed to eliminate the need for graph readers to apply certain processing components.…”