The Evolution of a SAS Database

Abstract

Initially, the concept of a Statistical Analysis System (SAS) database was explored as an alternative to the somewhat cumbersome system available at the time, one which required the statistician/report-programmer to access an Inquire database, create save files, and then create a SAS dataset for use in SAS table generation. In an attempt to eliminate some of these steps, a test situation was devised in which a SAS database was created from the same tape files which created the Inquire database and thus provided a parallel database. Editing of data continued to be done via the front-end preprocessor for the Inquire system, after which the SAS database was created for statistical programming use. Whereas the initial design allowed for the most efficient storage of data because the merging of all data occurred only temporarily at the time of report-generation, the major drawback was the amount of time required for the link program to run, especially in an interactive environment. However, the SAS database was so successful for SAS table programming that it has been continued with a second-generation database with one major design change. Instead of storing the individual datasets and linking them at the time of table-generation, the link process was moved back into the database creation. The SAS database now consists of a Subjects dataset, a combined Subjects, Efficacy, Visits, and Laboratory dataset, a Comments dataset, and a Normals dataset. SAS databases have now been created for several multicenter studies following this process, as well as for several single-center studies. In the case of the single-center studies, no front-end processing is done and data are read into the SAS database from sequential disk files. This SAS system, along with creation of SAS format libraries, has enabled the statistician/report-programmer to use a SAS database both in traditional batch processing and in a fully interactive environment.