A COMPARATIVE STUDY OF END USER COMPUTING PERFORMANCE FOR OBJECT-ORIENTED, EXTENDED ENTITY-RELATIONSHIP, AND RELATIONAL DATA MODELS AND QUERY LANGUAGES

Abstract

This study investigated the 1) effects of core OO, EER, and relational data models on user performance in data modeling task; and 2) effects of Orion-based OQL, EER-based KQL, and SQL query languages on user performance in query writing and query reading tasks through a laboratory experiment involving a total of 74 student subjects. These subjects were randomly assigned to three different groups. The study provides strong evidence that data models and query languages based on higher level abstractions have a significant impact on user performance. The findings support the belief that data models which have higher level abstractions (OO and EER models) are better than those which have lower level abstractions (relational model). In summary, OO model subjects performed significantly better than relational model subjects in terms of entity, generalization, unary one-one, binary one-one, and binary many-many relationships facets. EER model subjects outperformed relational subjects on generalization, binary many-many relationship, binary many-many relationship with relationship attribute, and ternary one-many-many relationship facets. However, there is no absolute superiority between OO and EER. For unary one-many relationship, binary one-one relationship, binary one-many relationship with relationship attribute, and binary many-many relationship facets, OO model is significantly better than EER. However, on both binary many-many relationship with relationship attribute and ternary one-many-many relationship facets, EER model is significantly better than OO model. The above results indicate. that EER model is more suitable for the representation of complex relationships while OO model is better suited for simple relationships. To improve novice user performance in data modelling, the advantages associated with OO model and EER model should be combined. In terms of time and accuracy, the exploration of three query languages found that higher level query languages based on higher level data models (OQL on OO model and KQL on EER model) led to better user performance than query languages based on lower level data models (SQL on relational model). The results suggest that OQL and KQL can lead to higher novice user productivity than SQL for query writing and reading tasks. Though no difference in user performance was detected between OQL and KQL for the query writing task, OQL was found to be easier to understand than KQL by the subjects for the query reading task. In summary, the experimental results suggest strong association between high abstraction in data models and query languages and novice user performance. Database vendors should increase the abstraction level of their data models and query languages to facilitate ease of use and understanding for novice database users.