The Effect of User View Characteristics on Database Design Performance of Novices

Abstract

Database design is a complex task for end users and novice designers. User views are an excellent input source for database design. This research investigates the effect of certain characteristics of user views on designer performance in the logical database design task. The study examines the extent to which performance of nonexpert designers engaged in normalizing user views and eliminating derived relationships can be predicted by examining the structure of user views. Two constructs that capture the complexity of a user view are defined - degree of nesting and derivation span. User views which varied on these parameters were included in a case. Subjects, who were enrolled in a database class and had learned the view decomposition approach, were asked to conduct logical database design of the case. Each view was graded using a predefined scheme. For each subject, there was a set of scores obtained for each user view. A multiple regression analysis using scores as the dependent variable indicated that both independent variables - degree of nesting and derivation span - significantly explained designer performance in normalizing the user views. Further, subjects had considerable difficulty in detecting derived relationships. The findings suggest that the degree of nesting and derivation distance cause complexity in database design tasks and lower designer performance.