This paper presents a metric of completeness in UML models and conducts an experiment (case studies) applying the metric. The first case study was done in the Image Processing domain. The authors obtained two UML diagrams from an Image Processing company. The second case study was done in the Embedded Controllers domain. Again two UML diagrams were obtained by the authors.
The authors then apply their metric for completeness to the UML diagrams they obtained. Below is a very brief outline of the topics discussed in the metric. Included aspects are: assessment of well-formedness rules, classes without method, interfaces without methods, abstract classes without sequence diagrams, messages without names, and messages without method.
The case studies showed significant variations in the conformance to and violations of the completeness rules. They also showed that different designers have substantially different amounts of occurrences of violations.