Guinnessometrics Against the Gold Standard: Randomization, Significance, and the Search for Validity

• Activity Number: 580
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Committee on Applied Statisticians
• Author(s): Stephen Ziliak*+
• Companies: Roosevelt University
• Keywords: W S Gosset aka Student ; Guinnessometrics ; Fisher ; Randomization ; Statistical Significance ; Economics

Abstract

What is Guinnessometrics? The statistical methods of William Sealy Gosset aka "Student" (1876-1937) are not as well known as they might be. Most statisticians have heard that Student's day job was Head Experimental Brewer at the Guinness Brewery. Yet few have heard how much Student's methods differed in reality from the methods that Fisher describes in Statistical Methods for Research Workers (1925) and elsewhere, under Student's name. Although Gosset pioneered the use of artificial randomization in the design of biometric experiments, and although he himself developed the small sample test of significance, three decades of experimental work and results convinced him to reject randomization and statistical significance in favor of balance and the economic approach. Repetition of his own experiments - at multiple sites and under varying conditions - gave, in addition, a better indication of external validity than any set of abstract assumptions, however mathematically elegant. Fisher, for his part, rejected Guinnessometrics, misleading others. A survey of best practice methods in some leading journals of science suggests that a little Guinnessometrics could help a lot.