An Experiment Comparing 2D and 1D Bar Graphs
MeasuringU
UX research and software supporting all stages of the user and customer experience.
The time it takes to click on an element on a page is a function of its size (Fitts’ Law).
People can only keep 7 ± 2 items in short-term memory (Miller’s Magic Number).
When multiple similar objects are presented, people will prefer the most different one (Von Restorff Effect).
People like to have laws named after them (Sauro’s Law).
There’s something compelling about succinct “laws” to help make design decisions. Following laws or conventions can help prevent designers from making certain mistakes.
And though making graphs is hardly the same design activity as building web applications, it’s still a designed interface that represents data. Ideally, having laws or rules for presenting data can reduce mistakes in interpretation and even prevent deception.
Some widely believed rules, however succinct, shouldn’t be strictly followed. For example, we showed how starting?graphs with 0?isn’t always a good idea. Context matters.
One such law, proposed by Edward Tufte several decades ago, stated:
“Volume increases at a higher rate than surface area. The number of information carrying (variable) dimensions depicted should not exceed the number of dimensions in the data”?(Tufte, 1983, p. 71).
This law (or at least rule or strong claim) from the literature on the visual display of quantitative data has been used to argue for the use of 2D over 3D graphs. In earlier articles, we reviewed the?literature on the use of 3D graphs?and reported our research on how quickly and accurately people could?judge the relative heights of 2D and 3D graphs. There are some exceptions, but in most cases, 2D graphs are as good as or slightly better than 3D graphs for accuracy of estimating values, accuracy of comparisons, and time to make decisions.
But what about 1D graphs? After all, area increases at a higher rate than height. If Tufte’s hypothesis about restricting the number of information carrying dimensions to the number of dimensions in the data, then wouldn’t it be better to show lines instead of columns in bar graphs?
In this article, we describe an experiment we conducted using the click testing feature of our MUIQ? platform to investigate how quickly and accurately people could compare two graphed values presented in 2D and 1D bar formats and which they preferred.
Discussion and Summary
Our analysis of 105 participants comparing 2D vs. 1D bar graphs found:
There was no difference in selection accuracy. Overall, selection accuracy was very high (> 98%). There were no significant differences in selection accuracy between 2D and 1D bar graphs.
领英推荐
It took participants longer to select 1D graphs. When there was a difference in bar height, the overall mean selection time per trial was about 1.4 seconds. Participants took just over a quarter of a second longer to select the 1D visualization, but much of this difference is likely due to Fitts’ Law effects (taking longer to click on smaller targets).
Selection was slightly slower for the smallest difference. Selection times were comparable for all differences in bar height except for the smallest (50% vs. 55%). Selecting the smallest difference took participants about 9% longer compared to the average of the other difference comparisons (1.51 vs. 1.39 seconds).
The location of the larger bar (left or right) did not significantly affect selection time. The difference in selection times was only 6 ms, so there was no evidence of a substantial time difference due to scanning from left to right.
There were no statistically significant differences in the control pairs. When there was no difference in bar heights, participants took over three seconds to click a selection. Overall, clicks were about evenly distributed between the left and right bars, with some participants clicking between the bars.
Participants preferred 2D graphs 10 to 1. The preferences were 72% for 2D, 21% no preference, and 7% for 1D. For those who had a preference, the ratio in favor of 2D was over 10:1.
Limitations and future research: One limitation of this research was that the task was very simple—selecting which bar was higher rather than, for example, having to estimate the magnitude of the height of one bar or the difference between the two. In future research on data visualization where the sizes of the visual representations being manipulated have very different sizes, it will be important to change the selection method to something like radio buttons to avoid the influence of Fitts’ Law.
Read the full article on MeasuringU's Blog
MUiQ Upcoming Feature: Moderated Studies
We are excited to announce that MUiQ will soon allow researchers to run moderated sessions directly in the platform.
With our?initial feature release, researchers will be able to:
Moderated Studies in MUiQ?will add a significant set of capabilities to further support every stage?of UX research.
Reach out today to learn more about MUiQ!