Information Extraction from Different Data Representation Forms on a CRT: Charts and Tables Janice M. Engberg F. Layne Wallace Dept. of Computer & Information Sciences 4567 St. Johns Bluff Rd., S. University of North Florida Jacksonville, FL 32224 USA Abstract. This paper studied the effects of four data presentation forms on users' abilities to extract information on computer screens. The results suggest that extraction time is shorter with bar graphs; however, tables were more accurate. 1. Introduction Information presentations using graphs and tables are often viewed as important tools for management as well as end- users. Although prior research has provided insight on the effects of presentation forms, findings are inconclusive and often conflict with each other. For example, some researchers indicate that information can be extracted more accurately from graphs than tables. While this may be true, other reports state that "graphs are no better than tables in presenting information" [Jarvenpaa85]. According to DeSanctis [cf. Jarvenpaa88], out of 29 studies, 12 indicated that tables were better than graphs, 10 found no significant difference, and 7 concluded that graphs were superior. These conflicting results have opened the door to further research in this area. Hoadley [Hoadley90] studied (1) the performance of subjects on extraction tasks using color versus monocolor information presentations, and (2) performance with different forms using the same colors. Subjects answered questions from four types of information presentations; bar graph, pie chart, line graph, and tabular chart. Hoadley found performance greatly decreased with the tabular format. Another pivotal study was done by Davis [Davis89] who presented financial data to subjects with four types of forms; line graph, bar chart, pie chart, and table. Test questions were rated on their level of complexity. Davis found that tabular charts resulted in performance superior or equal to that of the graphical charts. Furthermore, good performance for graphical charts was limited to questions of intermediate complexity. Benbasat, Dexter, and Todd [Benbasat86] presented data to subjects in the form of line graphs and tabular charts. The results from this study indicate that tabular formats are better for determining exact data values for computational purposes and graphical presentations are better for determining directions in the search for an optimal solution. Despite the seemingly contradictory findings, these studies provide a framework for further research. The purpose of this paper is to provide additional research in the area of human-factors, particularly with the effects of data presentation form on information extraction tasks. 2. Experimental Framework The objective of the current study was to examine the effects of presentation form on decision making tasks as defined by information extraction. The experimental framework consisted of a 3 x 2 (presentation form x pattern) factorial design which provided for controls and quantitative measurement. The independent variables were presentation form and color texture, solid and cross-section pattern. The research task used an information set and group of questions supported by prior research [Hoadley90, Davis89]. The questions were taken from Hoadley [Hoadley90] and Davis [Davis89]. The dependent variables were response accuracy and task completion time. The experiment was conducted on PC's with 14-inch monitors running MS-DOS and the Microsoft Windows 3.1 environment. Seven applications were written to accommodate the presentation forms (solid bar chart, pattern bar chart, solid line chart, pattern line chart, solid pie chart, pattern pie chart, and table). Each application was programmed to capture completion times and user responses. Brightness and contrast were directly controlled by the subjects. The presentation forms were displayed in four types; bar, pie, line, and tabular. Each presentation form represented a set of time-series data because of its popular business use [Hoadley90]. Pie charts, not well-known for their use with time-series data, were used because of their extensive use in business reports [Davis89]. The experiment consisted of subjects working through a series of 10 questions appearing on a CRT. Each question pertained to the same set of time-series data in either graph or tabular form. Individual question timings were added together to get a total completion time. Correct responses were added together to get a total accuracy score for each subject. Prior to viewing the presentation material, subjects were asked to provide demographic information and were tested for color blindness. If the results indicated that the subject was color impaired, his/her resulting data was excluded from the analysis. 3. Results of Data Analysis Seventy-one volunteers between the ages of 20 and 39 participated in the experiment (47 males and 24 females). Subjects were randomly selected for each group. In the current experiment 3 out of 50 males were color blind. According to prior research [Benbasat86], this is within the limits of normal expectancy. Data from an additional two subjects was removed because their vision was not 20/20 (either natural or corrected). An analysis of variance (ANOVA) was done to determine if a significant difference existed with completion times for graphic versus tabular charts. Table 1 illustrates the ANOVA test results. +------------------------------------------------+ |Source SS DF MS F P | +------------------------------------------------+ |Grp 118.757 6 19.7929 3.29 .0070 | |Error 384.986 64 6.01541 | +------------------------------------------------+ Table 1: ANOVA results for Graphs vs. Table completion times A Student-Newman-Keuls test indicated that (1) bar charts took significantly less time to complete than tabular charts, (2) no significant difference existed between tables, pie, and line charts, and (3) no significant difference existed between pie, line, and bar charts. This suggested that subjects took significantly longer with tabular charts (9.811 minutes) than with solid bar charts (6.061 minutes). An analysis of variance was done to determine if a significant difference existed with the accuracy of graphs versus tabular charts. Table 2 illustrates the ANOVA test results. +-----------------------------------------------+ |Source SS DF MS F P | +-----------------------------------------------+ |Group 54.4227 6 9.07046 8.13 .0001 | |Error 71.4363 64 1.11619 | +-----------------------------------------------+ Table 2: ANOVA results for Graphs vs. Tables response accuracy A Student-Newman-Keuls test indicated that (1) pie charts were significantly less accurate than bar, line, and tabular charts, (2) bar, line, and tabular charts were not significantly different. Tabular charts result in significantly higher scores than pie, line, and bar charts. Pie charts yielded significantly lower scores (5.900) than line (7.100), bar (7.761), and tabular charts (8.300). An analysis of variance was done to determine if a significant difference existed with completion times for the different graph types (bar, line, and pie). Table 3 illustrates the ANOVA test results. +---------------------------------------------------+ |Source SS DF MS F P | +---------------------------------------------------+ |Group 65.4896 2 32.7448 5.38 .0073 | |Texture 0.32011 1 0.32011 0.05 .8195 | |Grp*Txt 5.92989 2 2.96494 0.49 .6170 | |Error 334.756 55 6.08648 | +---------------------------------------------------+ Table 3: ANOVA results for Graph Type completion times A Student-Newman-Keuls test indicated that pie and line chart tasks took significantly longer to complete than bar charts. It suggested that information can be extracted significantly quicker from bar charts (6.0619 minutes) than line charts (7.99 minutes) and pie charts (8.211 minutes). An analysis of variance was done in order to determine if a significant difference existed between the accuracy of bar, line, and pie charts. Table 4 illustrates the ANOVA test results. +-----------------------------------------------------+ |Source SS DF MS F P | +-----------------------------------------------------+ |Group 36.32818 2 18.16409 15.29 .0001 | |Texture 1.661361 1 1.661361 1.40 .2421 | |Grp*Txt 0.411799 2 0.205900 0.17 .8413 | |Error 65.33634 55 1.187934 | +-----------------------------------------------------+ Table 4: ANOVA results for Graph Type response accuracy A Student-Newman-Keuls test indicated that scores for bar and line charts were significantly more accurate than those with pie charts, and there was no significant difference between bar and line charts. Bar and line charts produced significantly higher scores (7.43) than pie charts (5.90). The ANOVA showed no interaction effect between group and texture. This suggests that patterns neither deter or enhance extraction ability. 4. General Discussion The basic findings from the current experiment suggest that information extraction performance is dependent on presentation form and the task at hand. The current experiment and most prior research [Hoadley90, Davis89, Lusk79] agree that graphs and tables are significantly different when it comes to response time. The current experiment showed that it took significantly longer to extract information from tables than graphical charts. Hoadley [Hoadley90], Davis [Davis89], and Lusk and Kersnick [Lusk79] report that information can be extracted more quickly from tabular charts than graphical charts. The current experiment showed that tabular charts produce more accurate results than graphs. Davis [Davis89], Lusk and Kersnick [Lusk79] reported results similar to the current experiment. However, according to Hoadley [Hoadley90], tables produce lower scores than graphs. When comparing response time for each graphical chart, the current experiment indicated that bar charts take significantly less time to extract information from than line and pie charts. These results conflict with both Hoadley and Davis [Hoadley90, Davis89]. Hoadley [Hoadley90] showed no significant difference with response time. Davis [Davis89] reported that bar charts take longer to extraction information from than pie and line charts. Most research studies to date have been one-shot experiments and additional research is needed to determine if learning plays a role in information extraction over time. Finally, question complexity regarding information extraction is worth further examination since prior research suggests that performance differs depending on presentation form and question type [Davis89]. References [Benbasat86] Benbasat, I., A. S. Dexter, and P. Todd, The Influence of Color and Graphical Information Presentation in Managerial Decision Simulation, Human-Computer Interaction, 2, (1986), pp. 65-92. [Davis89] Davis, Larry R., Report Format and the Decision Maker's Task: An Experimental Investigation, Accounting, Organizations and Society, 14(5/6), (1989), pp. 495-508. [Hoadley87] Hoadley, E. D., and A. M. Jenkins, The effects of color on performance in an information extraction task using varying forms of information presentation, IRMIS Working Paper #W713, Indiana University, October, 1987. [Hoadley90] Hoadley, Ellen D., Investigating the Effects of Color, Communications of the ACM, 33(2), (1990), pp. 120-125. [Jarvenpaa85] Jarvenpaa, S. L., G. W. Dickson, and G. DeSanctis, Methodological Issues in Experimental IS Research: Experiences and Recommendations, MIS Quarterly, (1985), pp. 141-156. [Jarvenpaa88] Jarvenpaa, S. L., and G. W. Dickson, Graphics and Managerial Decision Making: Research Based Guidelines, Communications of the ACM, (1988), pp. 764-773. [Lusk79] Lusk, E. J, and M. Kersnick, The Effect of Cognitive Style and Report Format on Task Performance: The MIS Design Consequences, Management Science, 23, (1979), pp. 787-798.