Software to Link Database Interrogation and Statistical Analysis

Summary

This paper discusses the extent to which database handling facilities might usefully be incorporated in statistical analysis software. The rectangular data matrix so fundamental to almost all statistical analysis packages can certainly be output as a report file resulting from database interrogation. The circumstances in which this intermediate file could usefully be eliminated are discussed and arguments are presented on which of the three database models — network, hierarchical, relational — might be considered most appropriate for statistical work.

For an integrated system, retrieval is the vital database function and essentially involves row-by-row extraction of the data matrix. Data manipulation languages developed so far are generally hosted in some high-level language, often by means of a preprocessor. This paper considers such a mechanism in combination with Anderson’s (1980) approach to preprocessor-driven software for statistical analysis and comments on the feasibility and usefulness of such a system.