Testing the validity of a statistical analysis software

I responded to a Linked-in posting today, about validating Minitab, since that has been an assignment I was given about 6 years ago.  At that time I was asked to validate both Minitab and Excel for use in statistical analysis for the Department of Energy site that I was working.  We were not allowed to use any software that had not had its caclulational accuracy validated.

Another statistician sent me to the NIST site.  The National Institute of Standards and technology (NIST) has created the statistical reference data sets. they are found at http://www.itl.nist.gov/div898/strd/index.html

There are a series of data sets that are provided for each primary analysis, such as regression and anova. The data sets are created to test the weaknesses in the computational algorithms. Multiple sets are provided for each type of analysis, where each stresses the computation to a greater extent. With each data set is a reference result that is accurate to a lot of significant figures.

The validation is tedious, but it will answer your questions. By the way, my evaluation from 6 years ago found that Minitab, SAS, and Excel all were equally performing for statistical analysis, even after all the articles that have been written about how bad Excel is at statistics.