An experimental comparison of structured analysis and object-oriented systems analysis methodologie

This dissertation experimentally evaluates claims made by object-oriented systems analysis methodology originators that the object-oriented approaches more naturally represent information systems resulting in better understanding by people using the models associated with these methodologies. This dissertation (1) presents a research framework based on the cognitive psychology literature that describes how various factors might influence recall of information system features and understanding of the relationships among these features; (2) compares systems analysis methodologies using a paradigm-neutral characterization of the modeling constructs that must be provided by any systems analysis methodology to describe the end-user's domain; and (3) describes an experiment conducted to examine recall and understanding and reports the results of this experiment.;This dissertation examines the functional and structure models associated with Structured Analysis, Object-Oriented Analysis, and the Object Modeling Technique. An evaluation was conducted to determine if the models associated with the three systems analysis methodologies are informationally equivalent. The modeling constructs provided by each of these methodologies were compared using a set of necessary modeling constructs. Based on this comparison, the three methodologies are informationally equivalent. Next, an experiment was conducted to evaluate the recall and understanding achieved by studying the models associated with the three methodologies. A primary goal of this analysis was to determine if the models are computationally equivalent. One major finding of this dissertation is that there were few differences in recall and understanding across methodologies. The Object Modeling Technique subjects did spend significantly more time studying the models. Since the performance of subjects was approximately equal across methodologies and the Object Modeling Technique subjects spent more time studying the models, this indicates that the Object Modeling Technique models, as a set, may be computationally less efficient. Also, we found that the semantic richness of the information system influenced recall and understanding of information system features. Finally, there were a number of covariates that significantly influenced recall and understanding.