Statistical Analysis with Microsoft Excel 2002

John D. McKenzier, Jr.

Math/Science Division

Babson College

Babson Park, MA 02457-01310

mckenzie@babson.edu

(781)239-4479(W)

(781)239-6416(FAX)

Keywords: algorithms, random number generation, software documentation, spreadsheet, statistical software.

Abstract:

The most used spreadsheet package in the world today is Excel. Instructors, researchers, and practitioners increasingly use it to perform statistical analyses at schools and in the workplace. Among the reasons they use this software are is cost, (essentially none), its spreadsheet features (e.g. Links between data and output), and it familiarity (due to its dominant position among computer users). Still there are some major concerns with using Excel for data analyses that are unknown to most users. At the 2001 Joint Statistical Meetings Jon Cryer and the author organized an extremely well attended panel entitled, " The Use of Microsoft Excel for Statistical Analyses". The first three panelists presented an overview of some of these concerns. For example, Bruce McCullough illustrated his concerns with Excel's numerical results by explaining the serious deficiencies with its random number generators. They were followed by panelist Colin Bell from Microsoft who acknowledged that the existence of problematic algorithms and documentation. This presentation will summarize these concerns and identify others not mentioned by the panel. Then it will explain whether Microsoft has eliminated any of these concerns with Excel 2002, its most recent release of this popular spreadsheet package in Office XP.