UML Class Diagrams: An Empirical Study of Comprehension

Lecture Notes in Computer Science

2014

Abstract. Practical experience suggests that usage and understanding of UML diagrams is greatly affected by the quality of their layout. While existing research failed to provide conclusive evidence in support of this hypothesis, our own previous work provided substantial evidence to this effect. When studying different factors like diagram type and expertise level, it became apparent that diagram size plays an important role, too. Since we lack an adequate understanding of this notion, in this paper, we define diagram size metrics and study their impact to modeler performance. We find that there is a strong negative correlation between diagram size and modeler performance. Our results are highly significant. We utilize these results to derive a recommendation on diagram sizes that are optimal for model understanding.

Section: "Good" Layout Of Uml Diagramsmentioning

confidence: 99%

On the Impact of Layout Quality to Understanding UML Diagrams: Size Matters

Lecture Notes in Computer Science

2014

“…All of these factors, though, seem to have a rather small impact individually, as existing results had only marginal statistical significance, and pointed to small effect sizes (see e.g. [32,31,29]). We believe that this may be a consequence of the rather small populations, and the specific stimulus samples used in experiments by Purchase and others.…”

Section: Diagram Layout Factorsmentioning

confidence: 89%

“…[32,31]) and also provide sources to justify and explain these criteria (cf. [29]). Eichelberger [6] also discusses these criteria at length, and shows how they can be used in the automatic layout of UML class diagrams.…”

Section: Diagram Layout Levelsmentioning

confidence: 99%

On the impact of size to the understanding of UML diagrams

2016

Softw Syst Model

Document VersionAbstract Background: Practical experience suggests that usage and understanding of UML diagrams is greatly affected by the quality of their layout. While existing research failed to provide conclusive and comprehensive evidence in support of this hypothesis, our own previous work provided substantial evidence to this effect, also suggesting diagram size as a relevant factor, for a range of diagram types and layouts. Aims: Since there is no generally accepted precise notion of "diagram size", we first need to operationalize this concept, analyze its impact on diagram understanding, and derive practical advice from our findings. Method: We define three alternative, plausible metrics. Since they are all highly correlated on a large sample of UML diagrams, we opt for the simplest one. We use it to re-analyze existing experimental data on diagram understanding. Results: We find a strong negative correlation between diagram size and modeler performance. Our results are statistically highly significant, and exhibit a very large degree of validity. We utilize these results to derive a recommendation on diagram sizes that are, on average, optimal for model understanding. These recommendations are implemented in a plug-in to a widely used modeling tool, providing continuous feedback about diagram size to modelers. Conclusions: The effect sizes are varying, but generally suggest that the impact of size matches or exceeds that of other factors in diagram understanding. With the guideline and tool, modelers are steered towards avoiding too large diagrams.

“…The main focus of previous work on UML diagram types and their layout has been with one of four aspects: diagram comprehension (d. [22,19] and/or user preference (d. [18,27]), automatic layout (d. [7,10,16,8,4]), or one of a variety of diagram inference tasks, e.g., program understanding based on visualizations (d. [29]), or the role of design patterns in understanding (cf. [22,23]).…”

Section: Related Workmentioning

confidence: 99%

“…[18,17]) and also provide sources to justify and explain these criteria (ef. [19]). Eichelberger [6] also discusses these criteria at length, and shows how they can be used in the automatic layout of UML class diagrams.…”

Section: Related Workmentioning

confidence: 99%

Diagram Size vs. Layout Flaws

Proceedings of the 10th ACM/IEEE International Symposium on Empirical Software Engineering and Measurement

2016

CONTEXT: Previously, we have defined the notion of diagram size and studied its impact on the understanding of UML diagrams. Subsequently, questions have been raised regarding the reliability and generality of our findings. Also, new questions arose regarding how the quality of diagrams could be defined, and how it interacts with diagram size. GOAL: We pursue three goals. First, we want to increase the validity of our research by analyzing a substantially larger data set than before. Second, we broaden the generalizability of our results by including two more diagram types. Our main contribution, though, is our third goal of extending our analysis aspects of diagram quality. METHOD: We improve our definition of diagram size and add a (provisional) definition of diagram quality as the number of topographic layout flaws. We apply these metrics on 60 diagrams of the five most commonly used types of UML diagram. We carefully analyze the structure of our diagram samples to ensure representativeness. We correlate diagram size and layout quality with modeler performance data obtained in previous experiments. The data set is the largest of its kind (n = 156).RESULTS: We replicate earlier findings, and extend them to two new diagram types. We provide an improved definition of diagram size, and provide a definition of topographic layout quality, which is one more step towards a comprehensive definition of diagram quality as such. Both metrics are shown to be objectively applicable. We quantify the impact of diagram size and quality on diagram understanding. CONCLUSIONS: The overall results of previous studies are confirmed, while our previous recommendations for creating better diagrams are revised and refined.