Graphics : theories & experiments

The primary justification of this research lies in the current thinking among graphics theorists and Management Information Systems researchers that different forms of information representation facilitate different types of tasks, and that it is the task characteristics which essentially influence performance with a given information presentation. Three experiments were designed to investigate hypotheses drawn from the literature testing the relative strengths and weaknesses of various graphical representations for answering a series of questions. Three forms of graph format were studied comprising bars, symbols, and lines. Time is the primary dependent variable of interest in this research. Accuracy is a secondary criterion. The tasks investigated involved the extraction of relationships among elementary classes of information depicted on various attribute components of time series data: (1) Dependent Variable (DV) component (namely, information on scale-values, level relationships, and trends); (2) Primary Independent Variable (PIV) component (namely, information on abscissa time period); and (3) Secondary Independent Variable (SIV) component (namely, information on dataset classification). Experiment 1 tasks involved the extraction of the DV scale-value (QT), DV level relationship (Q2), and DV trend (Q3) based on specific time period information on the PIV component. Results indicated that lines took longest for Q1 when compared to bars and symbols. Conversely, experiment 2 tasks involved the extraction of time period information based on a specific DV scale-value (Q1), the DV level relationship between two points (Q2), or the DV trend among several points (Q3). No statistically significant time differences were found among the various graph formats. However, lines were less accurate to use than bars for answering Q1. Experiment 3 tasks involved the extraction of dataset information from the SIV component based solely on a specific DV scale-value (Q1), the DV level relationship between two points (Q2), or the DV trend among several points (Q3). Results revealed that the time required for answering either Q2 or Q3 was longest with bars. Together, these results strongly indicated that the degree of support provided by a particular graph format for a particular task is heavily dependent upon the matching of task characteristics with graph format characteristics. Having information related to either the answer or question anchored on the x-axis and/or y-axis was found to influence task performance with the different graph formats investigated. Also, information complexity of graphics was found to be a function of time periods and/or datasets. There was only partial evidence to suggest the influence of individual characteristics on performance.Business, Sauder School ofGraduat