Academic journal article Human Factors

Guidelines for Presenting Quantitative Data in HFES Publications

Academic journal article Human Factors

Guidelines for Presenting Quantitative Data in HFES Publications

Article excerpt

Editor's Note: This article was prepared at my invitation following a discussion with the editorial board over a rather ironic situation that we face in reviewing manuscripts. Whereas one would expect a discipline that claims special expertise in information display to be exemplary in presenting its own graphic material for publication, we find quite the contrary. The illustrations accompanying the typical manuscript submitted to Human Factors border on the abysmal. So I decided to approach several well-known researchers in this topic area with the idea of distilling their expertise into a set of useful guidelines. They agreed. However, for the sake of consistency, I felt it necessary to have the manuscript reviewed, and it was - by the HFES Communications and Publications Subcouncil. The end result is published here for the benefit of all would-be authors. Excerpts will be included in future editions of the Authors' Guide.

INTRODUCTION

The communication of scientific research requires decisions about how to present quantitative data (e.g., Loftus, 1993). Many options are available for presenting data, especially with the increasing capabilities of wordprocessing, spreadsheet, and graphics software for creating tables and graphs. However, having all of these options may complicate decisions concerning the communication of the data. This article attempts to help researchers to sort through the options by distilling the research and state-of-the-practice literature into a set of guidelines.

A concise statement of the philosophy of human factors is "know thy user." One way that this translates into specific design decisions concerning graphs is that designers should understand the tasks in which readers engage when they look at the displays. Users' tasks require certain sensory, perceptual, and cognitive operations (e.g., Gillan & Lewis, 1994; Lohse, 1993; Pinker, 1990; Simkin & Hastie, 1987). For example, a well-designed line graph can ease the discrimination of differences in the slopes of the lines (or of the differences in the angles among lines), which is valuable if the user's task involves identifying statistical interactions. However, if the user's task involves adding the values of conditions at various points, then discriminating among slopes would be of little value. In other words, a well-designed graph achieves a correspondence between the task demands and the operations the graph affords.

The philosophy that underlies the guidelines presented here extends the maxim "know thy user" to "know your users' tasks" and "know the operations supported by your displays." Finally, knowledge of tasks and display-supported operations can be integrated into "match the operations supported by your display to those the user needs to perform the task" (see also Larkin & Simon, 1987).

Deciding How to Present Quantitative Data

Making decisions about how to present data involves a multistage process, as illustrated in Figure 1. The first decision concerns the amount of data being presented. When presenting a small amount of data, authors should weigh the communication benefits of tabular or graphical presentation against the reader's cognitive costs (e.g., processing the data display and integrating the information from the display and text). Typically, a few data points that have a simple relation, such as one value being substantially greater than a second, would not require a graphical display for the reader to apprehend the difference; a table that simply showed the numbers would provide little communication benefit over listing those same numbers in the text. Consequently, those few data points might best be presented in the body of the text. In contrast, certain relations, even in a small amount of data, can be communicated more quickly and clearly by means of a graph. For example, if a 2 x 2 factorial experiment produced an interaction in the data, that interaction might best be communicated with two nonparallel lines in a graph, even though the lines would contain only two mean values each. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.