Designing science graphs for data analysis and presentation: the bad, the good and the better

Abstract

Graphs use spatial arrangement on the page or screen to convey numerical information: they are often easier to interpret than repetitive numbers or complex tables. The assumption seems to be made that creating good graphs is easy and natural: yet many bad or sub-optimal graphs encountered in the literature disprove this. We aim to help you to produce good graphs, and to avoid falling into traps that common computer software packages seem to encourage. A scientific culture is one where good graphs, and innovative and specialised approaches, are valued. Hence we explain some relevant psychology behind the interpretation of graphs. We then review a variety of graph formats, including some less common ones. Their appropriate uses are explained, and suggestions are given for improving the visual impact of the message behind the data while reducing the distraction of non-essential graphical elements. We argue against the use of pie charts and most three- dimensional graphs, prefer horizontal to most vertical bar charts, and recommend using box plots and multipanel graphs for illustrating the distribution of complex data. The focus of this publication is on preparing graphs for written communications, but most principles apply equally to graphs used in oral presentations. The appendices illustrate how to create better graphs by manipulating some of the awkward default settings of Microsoft Excel ( 2002 version) and illustrate the S-PLUS programming language (both programs are currently available on the computer network of the New Zealand Department of Conservation).