I have a dream: that professionals in all areas—business; government; the physical, life, and social sciences; engineering; medicine; and others—will increasingly use statistical experimental design to better understand their worlds and to use that understanding to improve the products, processes, and programs they are responsible for. To this end, these professionals need to be inspired and taught, early, to conduct well-conceived and well-executed experiments and then properly extract, communicate, and act on information generated by the experiment. This learning can and should happen at the undergraduate level—in a way that carries over into a student’s eventual career. This text is aimed at fulfilling that goal.

Many excellent statistical texts on experimental design and analysis have been written by statisticians, primarily for students in statistics. These texts are generally more technical and more comprehensive than is appropriate for a mixed-discipline undergraduate audience and a one-semester course, the audience and scope this text addresses. Such texts tend to focus heavily on statistical analysis, for a catalog of designs. In practice, however, finding and implementing an experimental design capable of answering questions of importance are often where the battle is won. The data from a well-designed experiment may almost analyze themselves—often graphically. Rising generations of statisticians and the professionals with whom they will collaborate need more training on the design process than may be provided in graduate-level statistical texts.

Additionally, there are many experimental design texts, typically used in research methods courses in individual disciplines, that focus on one area of application. This book is aimed at a more heterogeneous collection of students who may not yet have chosen a particular career path. The examples have been chosen to be understandable without any specialized knowledge, while the basic ideas are transferable to particular situations and applications a student will subsequently encounter.

Successful experiments require subject-matter knowledge and passion and the statistical tools to translate that knowledge and passion into useful information. Archie Bunker, in the TV series, All in the Family, once told his son-in-law (approximately and with typical inadvertent profundity), “Don’t give me no stastistics (sic), Meathead. I want facts!” Statistical texts naturally focus on “stastistics”: here’s how to calculate a regression line, a confidence interval, an analysis of variance table, etc. For the professional in fields other than statistics, those methods are only a means to an end: revealing and understanding new facts pertinent to his or her area of interest. This text strives to make the connection between facts and statistics. Students should see from the beginning the connection between the statistics and the wider business or scientific context served by those statistics.

To achieve this goal, I tell stories about experiments, and bring in appropriate analyses, graphical and mathematical, as needed to move the stories along. I try to describe the situation that led to the experiment, what was learned, and what might happen after the experiment: “Fire the quality manager! Give the worthy statistician a bonus!” Experimental results need to be communicated in clear and convincing ways, so I emphasize graphical displays more than is often done in experimental design texts.

My stories are built around examples in statistical texts on experimental design, especially examples found in the classic text, Statistics for Experimenters, by Box, Hunter, and Hunter (1978, 2005). This “BHH” text has been on my desk since the first edition came out. I have taught several university classes based on it and have incorporated some of its material into my introductory statistics classes. Most of the examples are simple at first glance, but I have found it useful to (shamelessly) expand the stories in ways that address more of the design issues and more of the what-do-we-do-next issues. I try to make the stories provocative and entertaining because real-life experimentation is provocative and entertaining. I want the issues and concepts to be discussable by an interdisciplinary group of students and the lessons to be transferable to a student’s particular interests, with enough staying power to affect the student’s subsequent career. An underlying theme is that it is subject-matter enthusiasms that give rise to experiments, shape their design, and guide actions based on the findings. Statistical experimental design and data analysis methods facilitate and enhance the whole process. In short, statistics is a team sport. This text tries to demonstrate that.

In 1974, I taught at the University of Wisconsin and had the opportunity to attend the renowned Monday night “beer seminars” in the basement of the late Professor George Box’s home. He would invite researchers in to discuss their work, and the evening would turn into a grand consulting session among George, the researcher, and the students and faculty in attendance. The late Bill Hunter, also a professor in the Statistics Department and an innovative teacher of experimental design, was often a participant. I learned a lot in those sessions and hope that the atmosphere of those Monday night consulting sessions is reflected in the stories I have created here. The other H in BHH is J. Stuart Hunter, also an innovator in the teaching of experimental design; his presentations and articles have influenced me greatly, and his support for this book is especially valued. He puts humor into statistics that nobody would believe exists. I attended several Gordon Research Conferences at which B, H, and H all generated a lot of fun. Statistics can be fun. I have fun being a statistician and I have tried to spice this book with a sense of fun. (Please note that this book’s title begins with fun.)

In this book, mathematical detail takes a backseat to the stories and the pictures. Experimental design is not just for the mathematically inclined. I rely on software to do the analyses, and I focus on the story, not formulas. Once you understand the structure of a basic analysis of variance, I believe you can rely on software (and maybe a friendly, local statistician) to calculate an ANOVA table of the sort considered in this text. Thus, I do not give formulas for sums of squares for every design considered. Ample references are just a quick Google or Wikipedia search away for the mathematically intrigued students or instructors so inclined. I give formulas for standard errors and confidence intervals where needed. I would be pleased if class discussions and questions, and alternate stories, led to displays and analyses not covered in my stories.

To offset my expanded stories, I limit the scope of this text’s topics to what I think is appropriate for an introductory course. I indicate and reference possible extensions beyond the text’s coverage. Individual instructors can tailor their excursions into such areas in ways that fit their students. This text can best be used by instructors with experience in designing experiments, analyzing the resulting data, and working with collaborators or clients to develop next steps. They can usefully supplement my stories with theirs.

Chapter-end assignments emphasize the experimental design process, not computational exercises. I want students to pursue their passions and design experiments that could illuminate issues of interest to them. I want them to think about the displays and analyses they would use more than I want them to practice turning the crank on somebody else’s data. Ideally, I would like for these exercises to be worked by two- or three-person teams, as in the real-world environment a student will encounter after college. (My ideal class makeup would be half statistics-leaning majors and half majors from a variety of other fields, and I would pair a stat major with a nonstat major to do assignments and projects.)

Existing texts contain an ample supply of analysis exercises that an instructor can choose from and assign, if desired. Some are listed at the end of this Preface. Individual instructors may or should have their own favorite texts and exercises. I would suggest only that each selected analysis exercise should be augmented by Analysis 1: Plot the data. These exercise resources are also useful windows on aspects of experimental design and analysis beyond the scope of this book that a student might want to pursue later in his studies or her career.

Software packages such as Minitab® also provide exercises. Teaching analysis methods in conjunction with software is also left to the individual instructor and campus resources. I use Minitab in most of my graphical displays and quantitative analyses, just because it suits my needs. Microsoft Excel® can also be used for many of the analyses and displays in this book. JMP® software covers basic analyses and provides more advanced capabilities that could be used and taught. Individual instructors should choose the software appropriate for their classrooms and campuses.

Projects provide another opportunity to experience and develop the ability to conceive, design, conduct, analyze, and communicate the results of experiments that students care about. I still recall my long-ago experiment to evaluate the effect of salt and sugar on water’s time to boil (not that boiling water was a youthful passion of mine, but getting an assignment done on time was). A four-burner kitchen stove was integral to the design. I cannot tell you the effects of salt and sugar on time to boil, but I was able to reject with certainty the hypothesis that “a watched pot never boils.” Again, I would encourage these projects to be done by small teams, rather than individually. Supplementary online material for the widely used text by Montgomery (2013) contains a large number of examples of student projects. I encourage students to seek inspiration from such examples. Much real-world research is motivated by a desire to extend or improve upon prior work in a particular field, so if students want to find better ways to design and test paper airplanes, more power to them. I also recommend oral and written reports by students to develop the communication skills that are so important in their subsequent careers. This is time well spent.

In-class experiments are another valuable learning tool. George Box, Bill Hunter, Stuart Hunter, and the Wisconsin program are innovators in this area. The second edition of BHH (Box, Hunter, and Hunter 2005) contains a case study of their popular paper-helicopter design problem. In my classes, I simplify the problem to a two- or three-factor design space to simplify the task and shorten the time required by this exercise.

This text provides in Chapter 3 enough of basic statistical concepts (estimation, significance tests, and confidence intervals), within the context of designed experiments, that a previous course in statistics should not be required. Again, I think that once concepts are understood, a student or working professional can understand and appreciate the application of those concepts to other situations. My hope is that this text will make it more likely that universities will offer an undergraduate (and beginning graduate)-level course in experimental design. This could be taught as a stand-alone course, or, as was the case when I taught at the University of Auckland, one course could have two parallel tracks: experimental design and survey sampling, taught by different instructors. This text should also be useful for short courses in business, industry, and government.

I am convinced that personal and organizational progress, and even national and global progress, depends on how well we, the people, individually and collectively, deal with data. The statistical design of experiments and analysis of the resulting data can greatly enhance our ability to learn from data. In George Box’s engagingly illustrated formulation (Box 2006), scientific progress occurs when intelligent, interested people intervene, experimentally, in processes to bring about potentially interesting events and then use their intelligence and the experimental results to better understand and improve those processes. My sincere hope is that this text will advance that cause.