Message from the president
By Joe Rodgers
Hello to the 1,278 members of Div. 5. This is my first presidential column, and I have lots to say, both to and about Div. 5. But I want my first column to define the topic I have chosen as my presidential focus for 2016 − 17: “Teaching Statistics and Quantitative Methods into the 21 st Century.” If it appears that I have forgotten the past 16 years, I have not. Our statistical and quantitative methods are outstanding and are becoming ever more so. But it does not seem that our teaching has entered the 21 st century yet. It is time to take some extraordinary measures to help our teaching methods catch up with our quantitative methods.
For many of us in quantitative (as well as qualitative) psychology, it all started with a stat class. But it did not end there.
I took math stat from Walter Kelley and psych stat from Jorge Mendoza, both in 1973 when I was an undergraduate at the University of Oklahoma. They felt like entirely different courses, and in a sense they were. Kelley's class had nothing to do with data analysis (or even with data for that matter). Mendoza's course, which used the statistics textbook by Robert McCall, was all about analyzing data. I took my first graduate statistics course from Larry Toothaker in 1975, using the textbook by William Hays, and then a second graduate stat course from Toothaker using Roger Kirk's textbook, “Experimental Design.”
For me, it most certainly did not end there. In the 40+ years since I was first introduced to the discipline of statistics, I have taught each of those classes, and around two dozen other courses related to quantitative methods, at six different universities. My most frequently taught course is introductory undergraduate statistics, which I have taught around 30 times, using approximately eight different textbooks. My “time series” for teaching this course started in 1980 and continues, though my awareness of the material in this course goes back to 1973. In all this time, the textbook material hardly has changed at all.
At APA in Denver this past August, I conducted a personal data analysis project on data analysis. I spent a couple of hours going from booth to booth, inspecting the tables of contents of the introductory undergraduate statistics textbooks from the major publishers. I have been studying such textbooks for years. Most are superbly written, with fascinating data examples. (I've recently used David Moore's text, which is difficult for a psychologist because of the dearth of measurement and design treatment, but it is outstanding for the dozens of fascinating datasets available in the book.) What I verified with my textbook inspection project, however, is that most tables of contents are interchangeable. Furthermore, and of deeper concern, they are almost identical to one in McCall's text from which I learned and taught so many years ago.
As the field of quantitative methods has progressed — as ANOVA has expanded into multilevel modeling, as factor analysis and GLM methods have combined into structural equation modeling, as new and innovative methods have emerged such as hazards modeling, mixture modeling and ARIMA modeling — the foundational courses that ultimately feed into those methods have changed almost imperceptibly. Tukey's exploratory methods and graphical procedures — stem-and-leaf diagrams, boxplots and scatterplot matrices — are in most modern statistics textbooks, along with more emphasis on effect size and confidence intervals (CIs). Notably, though, early versions of those graphs, and treatment of CIs, were present in 1970s textbooks.
I can make an argument that this is exactly the way it should be, but only to a point. The foundation, the starting point, should not change with the changing discipline. Understanding probability theory, random assignment, sampling distributions and the central limit theorem are still foundational topics, required entry material for later statistics and methods courses. But much of the basic knowledge that emerges from our introductory statistics textbooks does not easily transfer in support of modern methodology. Consider the following statements:
- “In testing and building models, small p -values are good, and big p -values are bad.”
- “An F -statistic is the ratio of two independent estimated variances, one that is sensitive to the treatment effect and one that isn't.”
- “Degrees of freedom is a measure of model complexity.”
Each of these statements is fundamentally correct in a certain restricted context; however, each also interferes with understanding modern modeling approaches. These and dozens of other concepts can be taught, at an introductory level, in much more modern and much more pedagogically sound ways.
Toward the goal of engaging a whole division (all 1,278 of you, as well as your families, friends and colleagues who are not in Div. 5) in careful scrutiny of our pedagogical approaches to teaching introductory methods classes (both undergraduate and graduate), and even in revising those methods in modern and progressive ways, we have the following events planned:
- Past President Scott Hofer established the inaugural Div. 5 mini-conference at the University of Victoria this past March, preceding the midyear Div. 5 Advisory Committee meeting. I plan to host a similar mini-conference at Vanderbilt in spring 2017 (also prior to the same midyear meeting). The title of the mini-conference, which we hope to stream live online, is: “ Teaching Statistics and Quantitative Methods into the 21 st Century.” We expect to invite and pay for a few teaching scholars to come and present talks on this topic. If you are interested in presenting such a talk, please contact me.
- This theme will continue at the 2017 APA meeting in Washington D.C. We will have symposia, paper sessions and poster sessions devoted to teaching and pedagogy (as well as other standard topics in quantitative and qualitative methods, as usual).
- My presidential address in D.C. will be devoted to modern teaching methods.
- Ultimately, we hope that at least one edited book will emerge from this presidential focus. At most, we hope for a complete transformation of teaching and textbooks for introductory methods, a transformation devoted to integrating teaching across the breadth of quantitative psychology.
A few scholars already have started fomenting such a revolution. At the risk of forgetting important contributors, I would like to give a shout-out to several who have influenced my own thinking and who have helped to get this “teaching transformation” started: Leona Aiken, Patricia Cohen, Patrick Curran, Harold Delaney, Lisa Harlow, Jeffrey Harring, Rick Hoyle, James Jaccard, Charles Judd, Rex Kline, Bud MacCallum, Scott Maxwell, Gary McClelland, Robert McGrath, Chip Reichardt, Robert Terry, Steve West and Keith Widaman.
If I inadvertently left your name off of that list, please be assured that your efforts are — and will continue to be — highly valued. If you are just irritated enough to be left off that list that you would like to help teach statistics and quantitative methods into the 21 st century, then welcome to the future. We look forward to seeing your name on the Div. 5 program next year in Washington, D.C.