A Set of Quality Criteria for Statistical Programming
- PDF / 393,380 Bytes
- 6 Pages / 504 x 720 pts Page_size
- 43 Downloads / 209 Views
0092-8615/2002 Copyright 0 2002 Drug Information Association Inc.
A SET OF QUALITY CRITERIA FOR STATISTICAL PROGRAMMING ALLANGLASER,MS Manager, Statistical Programming, Merck 8c Co., Inc., Blue Bell, Pennsylvania
The analysis and reporting of clinical trial data involves a substantial amount of statistical programming. Although different organizations may take different approaches to accomplish programming tasks, there are common criteria upon which the quality of the work may be evaluated. This paper will identify a set of criteria, comment on the importance of each, and offer practical guidance on objective measurement. Key Words: Statistical programming; Quality; Quality criteria
INTRODUCTION
THE STATISTICAL PROGRAMMING for a clinical trial generally involves both biostatisticians and programmers, and includes complex manipulation and restructuring of safety and efficacy data, reporting of these manipulated data, and constructing tools that will be used for the review of these data. While the statistical analysis follows a welldefined path, the programming effort may not be as well established. There is a common perception that this type of programming is at least as much art as it is science, and that any programming oversights will be captured during a later quality control step. This perception is exacerbated by the lack of formal computer science education of many practitioners. Identification of quality criteria for programs, and giving attention to these criteria during the programming activity, are important. Relying upon sources outside of the pharmaceutical industry, relevant criteria can be found. Although there is a wealth of literature expounding upon these criteria, the fact Reprint address: Allan Glaser, Merck & Co., UN A102, 785 Jolly Road, Blue Bell, PA 19422.
remains that the measurement of many of them remains subjective. As the programming art advances, and as new tools and approaches become available, this situation is likely to change, albeit slowly. For the foreseeable future, it is appropriate to highlight these criteria, emphasize their importance during the normal course of business, and encourage the routine use of tools and techniques that tend to maximize quality.
THE CRITERIA This paper will present and briefly discuss 11 criteria. These represent an aggregation of established criteria for applications programming, criteria identified in computer science literature, and criteria suggested by selected government and private agencies (1). Where appropriate, methodologies that support these criteria and tools and techniques for measuring and evaluating conformance to these criteria will be given. These criteria are not disjoint; there is clearly substantial overlap. These criteria are not intended to be all-inclusive, and the needs of an organization must dictate their relative importance as well as the possible inclusion of other criteria.
565
Downloaded from dij.sagepub.com at Bobst Library, New York University on May 8, 2015
Allan Glaser
566
Correctness
A program is deem
Data Loading...
