A Statistical Database Model : Its Uniqueness and the Design Procedure

概要

論文の詳細を見る
Existing statistical tables show many varieties in representing the same fact. If we construct statistical databases without prohibiting various representations of the same fact, the logical design will change from one designer to another. Thus, it is difficult to perform data interchange and share the statistical data. In order to represent the components uniquely, we clarify the concepts of pure summary attribute and primitive category attribute in the statistical database model. A pure summary attribute cannot mix any component of category attributes. A category of a primitive category attribute should not be composed of two or more categories of other category attributes. The principle behind the model is "One fact in one way". Based on the proposed model, starting from existing statistical tables, we present a design procedure. We first perform purification and orthogonal decomposition which decompose existing statistical tables into unique atomic representations, and then synthesize them into canonical form of original tables. The canonical form is not always unique. However through this procedure, the point that causes non-uniqueness of the statistical database design can be identified. The authors hope that this study can be of use for the possibility of standardizing the logical design in statistical databases.
一般社団法人情報処理学会の論文
1989-08-30