A Standardized Form for Describing Programs, Packages and Systems for Statistical Applications

Abstract

This article is intended to stimulate exchange of information on computing software for statistical applications. It describes a standardized form developed by the ISI Committee on Statistical Computation for identifying and describing statistical computing capabilities. It is hoped that the form will provide a basis for compiling an information bank on programs, packages and systems for statistical applications. Recent years have seen the steady evolution of a more integrated approach to statistical computing. One of the earliest and perhaps the most widely distributed statistical program packages was the BIMED series developed at U.C.L.A., followed by the recently distributed version (BMDP, 1975). A more recent stage of evolution is represented by the development of high-level language systems such as GENSTAT (1975), P-STAT (1971), SAS (1976), SPSS (1975) and STATJOB (1973), which aim to provide general processing capability for a wide range of computing tasks. At a lower level of integration there are a number of welldocumented subroutine libraries for statistical applications, e.g. IMSL (1976). Dissemination of information about available packages, libraries, processing systems and languages is becoming increasingly important, but has received only limited attention so far. Some journals (e.g. Communications of the ACM and Applied Statistics) publish computing algorithms. Articles surveying various aspects of statistical computing have appeared in several journals (for example, ACM, JASA, Technometrics; see also the work of the ASA Committee on Statistical Computation (1971, 1974a, 1974b) for additional references. However, in general, the current scientific journals are either unsuited or unprepared to act as a repository for other than brief descriptions of large-scale packages and systems. Furthermore, the dissemination of information that now takes place is largely restricted to the highly developed industrialized countries. It is therefore desirable that one or more international information centres be formed to maintain a directory of statistical computing resources, based on a standardized description of them in terms of capabilities, supporting documentation, availability for distribution, and implementation requirements. On a longer term basis it is hoped that such centres would also provide information on the extent of certification and/or performance evaluation of the documented resources. An important objective in establishing one or more information centres is to avoid unnecessary duplication in the development and maintenance of statistical software. Another objective is to foster higher standards of design, implementation, documentation and supporting services. These standards should foster greater portability (transferability) of software between different computing environments. It is hoped that the increased availability of information and the increased portability of statistical software will be of direct benefit to statisticians in all countries.