Do Learners to Create an Artifact with Good Quality Make a Number of Trials and Errors during the Editing Process?

Courses to train modeling competence are held in higher educational institutions. These courses focus on the final artifacts submitted by learners for analysis and/or evaluation and do not clarify the intermediate creation process. This study focuses on quality of the final class diagram submitted by learners and trials and errors during the creation process. We tried to verify a hypothesis that learners to create an artifact with good quality make a number of trials and errors during the editing process. We collected the fine-grained data regarding the editing history of class diagrams created by learners by our tool called KIfU. We verified the hypothesis by analyzing the collected data. First, we defined two metrics ("Number of editing" and "Number of movement" of symbols in the artifact) calculated from the data that show the learners´ trials and errors. Second, we divided learners into two groups, group that created artifacts with good quality and that with not good quality. Finally, we compared the metrics between the groups. The results show that there are statistically significantly differences of values of both metrics between the groups. We conclude the hypothesis we set is correct from the result.