Page 204 - Statistics II for Dummies

188 Part III: Analyzing Variance with ANOVA

The secret lives of statisticians

Sometimes it’s hard to imagine famous people He wrote many papers, two of which were
having real lives, and it may be especially hard so important that they made it onto the list
to picture statisticians doing anything but sit- of the top 25 most-cited statistical papers of
ting in the back room calculating numbers. But all time.
the truth is, famous statisticians are interesting ✓ William Sealy Gosset, or “Student”: The
folks with interesting lives, just like you and me. first name included on the Student Newman-
Consider these stellar statisticians:
Keuls test is a story in itself. “Student” is
✓ Henry Scheffe: Scheffe was a very dis- a pseudonym of the English statistician
tinguished statistician at University of William Sealy Gosset (1876–1937). Gosset
California, Berkeley. One of his five books was a statistician working for the Guinness
The Analysis of Variance, written in 1959, is brewery in Dublin, Ireland, when he
the classic book on the subject and is still became famous for developing the t-test,
used today. (I used it in grad school and still also known as the Student t-distribution
have a copy in my office.) Scheffe enjoyed (see Chapter 3), one of the most commonly
backpacking, swimming, cycling, read- used hypothesis tests in the statistical
ing, and music, having learned to play the world. Gosset devised the t-test as a way
recorder during his adult life. Sadly, he died to cheaply monitor the quality of beer. He
from a bicycle accident on his way to the published his work in the best of statistical
university in 1977. journals, but his employer regarded his use
of statistics in quality control to be a trade
✓ Charles Dunnett: Nicknamed “Charlie” secret and wouldn’t let him use his real
(did you ever think of famous statisticians name on his publications (although all his
as having nicknames?), Dunnett was a dis- cronies knew exactly who “Student” was).
tinguished, award-winning professor in the So if not for Guinness beer, the Student’s
Departments of Mathematics, Statistics, t-test would have been called the Gossett
Clinical Epidemiology and Biostatistics at t-test (or you’d be drinking”Gosset beer”).
McMaster University in Ontario, Canada.

Going nonparametric with
the Kruskal-Wallis test

The Kruskal-Wallis test was developed in 1952 by American statisticians
William Kruskal (1919–2005) and W. Allen Wallis (1912–1998). The Kruskal-
Wallis test is the nonparametric version of a multiple comparison procedure.
Nonparametric procedures don’t have nearly as many conditions to meet
as their traditional counterparts. All the other procedures described in this
chapter require normal distributions from the populations and often the
same variance as well.

16_466469-ch10.indd 188 7/24/09 9:41:45 AM