Development of a valid and robust metric for measuring question complexity in computer graphics experimentation
Theophilus B. A. Addo, A. Milton Jenkins · Indiana University eBooks · 1989
The objective of this research is to develop and validate a metric for measuring question complexity in computer graphics experiments. A rudimentary theory proposed by Jacques Bertin identifies question complexity as one of the variables that affect the performance of tasks involving the extraction of information from graphical presentations. The information extraction phase of managerial decisions involving the use of graphical presentations has traditionally been ignored in prior graphics research. Two important reasons for this are the following: (a) the lack of theory which could identify, as well as specify the relationships among, the important variables involved in the use of graphical presentations; and (b) the lack of valid metrics for measuring and controlling the levels of these variables in experiments. The question complexity metric developed in this research was validated using a laboratory experiment. The mathematical formulation of the metric was used to guide the development of a set of questions which was answered by the subjects using three forms of graphical presentation: line graph, pie chart, and bar chart. The experiment was designed as a factorial two-way analysis of variance; the two factors being Question and Form of Presentation. The dependent variables measured were Time (the time taken to extract the answers to the questions from the presentations) and Accuracy (the accuracy of the answers extracted). The analysis of the data collected showed that the metric is both valid and reliable. The metric is composed of three factors. The relative importance of these factors was determined using multiple regression analysis. In decreasing order of importance, the metric's component factors are: image, step, and polarity.