A statistical user interface for the Relational Model of data

Abstract

In the decade since the introduction of the Relational Model as a user view of large stored data bases, a variety of user languages have been proposed and a number of experimental systems have been implemented. The current computer science literature is replete with papers on the theoretical and practical aspects of the Relational Model and its implementation, as are most recent texts on data management systems.Implicit in the design of the user languages of most database systems, including those based on the Relational Model, are assumptions regarding the patterns of access to and the usage of the content of the database. Somewhat oversimplified, the assumed pattern of access is to search for a particular occurrence (case, observation) in the database which satisfies a given condition, and then to display the values of all attributes (fields, variables) of that one occurrence. The languages are designed to permit users to pose queries such as, for example, "what widgets do we buy from ABC industries?" or "display Jones' employment history". Queries of this type are termed informational queries; and systems supporting such queries with appropriate user languages and internal data storage techniques and access methods are information systems .A statistical query, similarly oversimplified, specifies a pattern of access to most, if not all, of the occurrences in a database, and a usage pattern of at most a few of the attributes. Examples of statistical queries are "what is the average size of our purchase orders?" and "display the number of employees by race, sex, and job category". Current statistical systems have limited capability for performing analysis over large and complex data collections, and their user languages reflects this limitation. A statistical query, as defined here, need not involve sophisticated mathematical analysis; the distinction between informational and statistical is derived from the antithetical patterns of access to and usage of the data content of a database. Most work on access languages for relationally based data systems has been on information query languages; very little work has been done on statistical query languages.This paper, then, discusses some elements of a language for statistical queries for a data system employing the Relational Model as the user view of large stored data bases.