Experimental analysis about the graphic presentation of data in line graph and bar chart under superposition and juxtaposition

Keywords:

charts, graphs, bar chart, line graph, superposition, juxtaposition, graphical perception, visual displays, graphical displays

Abstracts

graphicexperiment 1

The influence of graphtype and graphstructure on the perception of trends

The presentation of many trendlines can be handled by:
- one chart for each trendline (Juxtaposition) or
- all trendlines in one chart (Superposition).
Both bar chart and line graph seem to be adequate graphtypes. The hypothesis was: The difference between the graphtypes grows on with increasing difficulty which was varied by the number of trendlines.

The aim of study consisted in the empirical testing of the influence of the following 3 factors:
- graphtype (vertical bar chart, line graph),
- graphstructure (juxtaposition, superposition),
- number of trendlines (1,2,4 or 8 trendlines).
The dependent variables were time and accurancy of perceiving sign respectively sequence of signs of slopes of a special trendline (eg. - + by an u-shape). 20 subjects participated on a complicated computer experiment.

Results: If the presentation contains only one trendline there is no difference between bar chart and line graph. This could be the reason for no difference between bar chart and line graph under juxtaposition independent of the number of trendlines. Only under superposition the difference between the graphtypes increases with the number of trendlines in favour of the line graph. (e.g.: result: reaction time to find a panel with a special sequence of signs of slopes) Up to 4 trendlines line graph under superposition achieves the same results as the two graphtypes under juxtaposition. But by 8 trendlines superposition leads to more errors than juxtaposition. This study is the first part of a project which will specify and test the interactions between tasks and presentations of charts.

graphicexperiment 2:

Experimental analysis about the perception of trends and trend comparisons in bar chart and line graph in superposition and juxtapositon

The broadening of the first graphic experiment consisted of the examination of the influences of graphtype (bar chart, line graph), graphstructure (superposition, juxtaposition) and number of panels (=trends) on the speediness and accurancy of the answer of special questions. The main interest was in the examination of such questions, which concern data in a panel as a whole and whose answer requires the comprehension of the relations in its entirety. This are questions about the trendtype of a data series, about differences of an expected trend or about similarities or dissimilarities of several trends. The advantages of a special graphic presentation should be primary achieved by a better perception.

To test the experimental conditions there was developed a complex computer program, whom underlied a generator for the construction of artificial data. So there was a big range of different data constellations available. 25 students took part in this experiment.

(You can download one special experiment of graphicexperiment 2: 'Identification of a trendtype in line graph and bar chart'. To run the experiment you need Toolbook 1.5 runtime.
for downloading !

All statistical results consistently verify clear advantages for the line graph by such questions (e.g. result: reaction time to find the panel with deviation from an ideal shape) .

graphicexperiment 3

Global comparison, local comparison and point reading in line graph and bar chart on condition of superposition and juxtapositon

In continuation of the 1994 by Jacobs started project this study deals with further questions that should be answered with graphical presentations. Bar chart and line graphs served as graphtypes, superposition and juxtaposition as graphstructures. The single graphs under juxtaposition were partly tested in vertical or horizontal order. The main focus of this study was, which of this graphic variants are best suitable for global comparison specially for the estimation of sums respectively means and variability of panels. In addition of that local comparison as well as point reading were examined for all graphic variants.

A complex program controlled the course of the experiment. For each subject the program generated other data by definite rules and so a variety of different data constellations could be provided. 25 students participated in the experiment free of charge.

For the comparison of sums respectively means of panels the prevailing advantage of the graph type depends on the graph structure. By the comparison of variabilities the line graph is clearly superior to the bar chart. To compare elements of different panels at a definite time point there is no difference between the two graph types; but there is a clear advantage of superposition in comparison with juxtaposition, which achieves practical importance. In all graph variants there is no difference in point reading.

graphicexperiment 4:

The perception of particular relations in bar chart and line graph under superposition and juxtaposition

This study is the end of the 1994 by Jacobs started project: "Experimental analysis about the graphic presentation of data in line graph and bar chart under superposition and juxtaposition". The main interest of the study is the perception of aspects that can be qualified as pecularities in a set of data. This includes questions about the largest or smallest value, about the largest difference between two sets of data, about the extreme slope within a panel or about the largest slope difference between several panels within a certain area. Under favourable constellations of data such pecularities hit right in the face but even under more difficult constellations the visual perception makes the search easier and leads to a faster decision as a tabular identification. The aim of the study was the empirical clarification of the matter: Which graphtyp (line graph, bar chart) seems in which graph structure (superposition, juxtaposition) particularly suitable for which question?

Like in the past experiments a computer program controlled the whole course of the experiment and provided a variety of different data constellations for testing. 46 students participated in the experiment free of charge.

The quantity of results confirms the expectation that the advantage of a graphic variant depends on the type of question. It could be verified clear effects of graph structure especially in favor of superposition, (e.g.: Results to "Where is the greatest difference between two panels?)., mediocre differences of graph types in both directions as well as numerous interactions between graph type and graph structure.

Some Experiments are now available in English (22.4.1997)

(Translation : Susanne Kroll)