statrefs home‎ > ‎Main‎ > ‎Software‎ > ‎

.Software Validation

Software Validation

Whatever statistical software package is used, it is a good idea to ensure that it gives results that are expected before relying on conclusions based on analyses performed using that software package. 
  • Regulatory Value:  In some industries, and depending on the risk involved (i.e. patient / customer safety), validating software for use may be a regulatory requirement.
  • Business Value:  Taking the time necessary to validate software for use has several benefits, to include: it can help to ensure that the software's algorithms are working properly; it can help the user of the software to understand how to properly use the software, to include knowing which options or menu choices to make to get the appropriate result.
The topic of software validation is too broad to cover here.  Applicable standards include the IEEE SWEBOK and ISO 12207.  However, for statistical software packages, there are several approaches that can be used.  These can include matching the software's output for a particular type of analysis against known good results (textbook, another validated software package, a standard set of output provided by a recognized body such as NIST).

IEEE Software Engineering Body of Knowledge (SWEBOK)

For verifying computational results, NIST offers NIST Statistical Reference Datasets with certified results.