Authors: Chua Eng Huang Cecil, Ee-Peng Lim, Roger H. L. Chiang
Tags: 1998, conceptual modeling
Many different kinds of algorithms have been developed to discover relationships between two attribute groups (e.g., association rule discovery algorithms, functional dependency discovery algorithms, and correlation tests). Of these algorithms, only the correlation tests discover relationships using the measurement scales of attribute groups. Measurement scales determine whether order or distance information should be considered in the relationship discovery process. Order and distance information limits the possible forms a legitimate relationship between two attribute groups can have. Since this information is considered in correlation tests, the relationships discovered tend not to be spurious. Furthermore, the result of a correlation test can be empirically evaluated by measuring its significance. Often, the appropriate correlation test to apply on an attribute group pair must be selected manually, as information required to identify the appropriate test (e.g., the measurement scale of the attribute groups) is not available in the database. However, information required for test identification can be inferred from the system catalog, and analysis of the values of the attribute groups. In this paper, we propose a (semi-) automated correlation test identification method which infers information for identifying appropriate tests, and measures the correlation between attribute group pairs.Read the full paper here: https://link.springer.com/chapter/10.1007/978-3-540-49121-7_3
