Fundamentals of Statistical Experimental Design and Analysis

1Introduction

Motivation: Why Experiment?

Statistics is “learning from data.” We do statistics when we compare prices and specifications and perhaps Consumer Reports data in choosing a new cell phone, and we do it when we conduct large-scale experiments pertaining to medications and treatments for debilitating diseases.

Much of the way we learn from data is observational. We collect data on people, products, and processes to learn how they work. We look for relationships between variables that may provide clues on how to affect and improve those processes. Early studies on the association between smoking and various health problems are examples of the observational process—well organized and well executed.

The late Professor George Box (Box, Leonard, and Wu 1983; Box 2006; and in various conference presentations in the 1980s) depicted history as a series of events, some interesting, most mundane. Progress happens when there is an intelligent observer present who sees the interesting event and reacts—who capitalizes on what has been learned. Box cited the second fermentation of grapes, which produces champagne, as an especially serendipitous observation. (Legend has it that a French monk, Dom Pérignon, made the discovery: “Come quickly, I’m drinking the stars!” (Wikipedia 2015).)

Now clearly, as Professor Box argued, progress is speeded when interesting events happen more frequently and when there are more intelligent observers present at the event—“more” in the senses of both ...