On the accuracy of statistical distributions in Microsoft Excel 97

The paper deals with the numerical accuracy of some statistical functions in Microsoft Excel (Excel version in Microsoft Office 97 for Windows). It is shown that the computation of some discrete distributions (Binomial, Poisson, Hypergeometric) fails even for probabilities in the central range between 0.01 and 0.99 and even for parameter values that cannot be judged as too extreme. Furthermore our tables show that the computation of some continuous distributions and their quantiles (Normal, Chi-square, F, t) is unsatisfactory in the tails of the respective distribution. In the view of the author a user of Excel may expect that the computed results are correct with all given digits; if Excel prints out a result with nine or ten digits and in fact only one or two of these many digits are correct, then such a result is unacceptable. As to the accuracy a user may expect of a statistical program see Knusel (1995). So one has to warn statisticians against using Excel functions for scientific purposes. The exact values in the tables were computed by means of SAS for Windows (1996), and they were checked with the program ELV in Knusel (1989). (SSNinCSDA 26, 375-379 (January 1998)).