What's Your Favorite Statistics Textbook?

Over the years, I have made use of a number of statistics textbooks. I have my personal favorites, the ones that remain on my bookshelf, but I find myself getting rid of most of the stats books that find their way into my hands, for a variety of reasons, but usually one of two: Books that are methodologically sound and statistically rigorous are very dense and are usually hard to read. Books that are easy to read are usually not rigorous.

The following modern statistics books seem to strike the right balance, and I have used both in graduate and undergraduate classes.

Gravetter, F., & Wallnau, L. (2008). Essentials of statistics for the behavioral sciences (6th ed.). Belmont, CA: Thomson.

Howell, D. (2008). Fundamental statistics for the behavioral sciences (6th ed.). Belmont, CA: Thomson.

I also refer often to Welkowitz, Cohen, and Ewen (2006). Although this book is a little harder to read than either Howell or Gravetter & Wallnau, Welkowitz, et al. has a good discussion on statistical power and effect size. I have had this book from the second edition published by Academic Press back in 1976.

Welkowitz, J., Cohen, B., & Ewen, R. (2006). Statistics for the behavioral sciences (6th ed.). New York: Wiley.

In addition to these modern books, some "classics" that I cannot part with are:

Hays, W. L., & Winkler, R. L. (1970). Statistics: Probability, inference, and decision. New York: Holt, Rinehart, and Winston.

Hays, W. L. (1973). Statistics for the social sciences (2nd ed.). New York: Holt, Rinehart, and Winston.

One of my all-time favorite statsistics books is:

Roscoe, J. T. (1975). Fundamental research statistics for the behavioral sciences (2nd ed.). New York: Holt, Rinehart, and Winston.

Roscoe gives the rationale and the computational approach to a variety of tests. His approach is so stratightforward, that I have built Excel statistics templates for many statistical tests by following his instructions and just translating them into Excel functions and formulas. This book is always in easy reach, and has a bunch of sticky notes for bookmarks.

And, of course, no statistician's library would be complete without Tukey's classic on exploratory data analysis:

Tukey, J. (1977). Exporatory data analysis. Reading, MA: Addison Wesley.

I tracked the Tukey book down on Amazon, and I consider the $40 I spent for it an excellent investment.

Unfortunately, those of us who teach online or as adjunct professors, and especially as online adjuncts, do not usually get to select the textbooks we use. I have had to teach from some miserable excuses for books. In one graduate research methods text, the author claimed that computing a t test was beyond the scope of the course! The same author later talked about Multiple R having both positive and negative values! I cringed. I found myself writing little tutorials for my students to correct the inaccuracies in their text.

What statistics books do you like?

Reporting Statistics in APA Format

There will be a new and much needed sixth edition of the Publication Manual of the American Psychological Association (APA) in a few weeks. Mine is already preordered. I have written extensively on reporting statistical results in APA format, something that is either not covered, or not covered well, in most online behavioral statistics courses. Of course, my earlier work referred to the fifth edition of the APA manual (APA, 2001), which is getting a little long in the tooth. I can't wait to get my hands on the new one, not just because of the statistics reporting instructions, but also to see the needed updates in the way to cite electronic sources.

I doubt that the new edition will contradict the old one in any significant way, though I will of course update my tutorials when the new edition is available. In the meantime, let me address some of the common mistakes students (and unfortunately many of my colleagues) make in writing about statistics, at least as far as the APA is concerned.

Italics or No Italics?

The APA manual requires that Roman letters used to represent sample statistics be displayed in italics. Thus, you italicize t, z, F, p, r, M (for mean), and SD (for standard deviation). Interestingly, the APA manual says NOT to italicize Greek letters used as symbols for population parameters or statistical indicators. Thus, you do not italicize α, β (used for regression coefficients and for the probability of Type II error), σ, η², or χ². I know common practice is to ignore these small details, but APA says to do it the way I describe.

Hyphens

Many statistics instructors (including yours truly in my own writing in the past) have made common practice of using forms such as t-test, F-test, F-ratio, z-score, and p-value as nouns. The APA manual says you should only use the hyphen when the combined form is used as an adjective to modify the following word. Thus, according to APA, you should write, "The results of t tests," but write "t-test results."

Leading Zeros

According to the APA manual, values that cannot exceed 1 should be reported without a leading zero. Thus we report p = .02, r = .87, and η² = .72. On the other hand, values that are less than 1 but which could exceed 1 should have the leading zero, so we would write z = 0.47, F = 0.93, and d = 0.69.

How many Decimals?

The APA manual says that in general well-constructed measurements can be reported to two decimal places. The manual says further that test statistics such as t, F, z, r, and χ² should be reported to two decimals. Probabilities and proportions should also be reported to two decimals in most cases, so that the minimum probability we would report in typical circumstances would be p < .01.

More on this later!

Helping Online Statistics Students

Teaching statistics online is a challenge, as is taking a statistics course online. When I teach statistics in the classroom, the students (at least the ones who came to class) and I are together in the same place. In addition to projecting my computer, I can use the board to illustrate and help students visualize the concepts and computations.

Online statistics courses are challenging for two main reasons. The immediacy of being in the same place at the same time is absent, and many "teachable moments" are thus lost. Although there are electronic whiteboards that can be integrated with the online classroom, they are not the same as a chalkboard or a dry erase board.

Here are some things I have found that help bridge the gap between the online student and the online professor of statistics. Mostly, I developed or learned these things through trial and error. Even though there are some books (some of them good) about teaching online, they are really more about strategies of teaching than they are about the mechanics of teaching.

Use a tablet PC. Although it is a rather esoteric device with only a small niche market, the tablet PC allows you to write on the screen with digital ink. The Tablet PC's commands sit on top of Windows XP or Vista, and allow you to annotate the screen much as you would draw or write on a chalkboard or whiteboard.

Use a screen capture program. You can make prerecorded "lecturettes" with the tablet PC and a program like Camtasia Studio. There are many other such programs, but I enjoy the flexibility that Camtasia offers. I can wear a headset and record the computer screen, my annotations, and my voice. Students may watch and listen as many times as they like. Students have commented that these lecturettes are very helpful.

Use a computer program to which your students have easy access. Not everyone has the same calculator, and most students do not have dedicated statistics software. But everyone has (or can easily get) Microsoft Excel. If you want free software, you can use the Calc program that comes with OpenOffice. Using a spreadsheet makes it easy to show the data, the formulas, and the results in the same place at the same time. Cognitive psychologists call this a "unified learning space." Used properly (and I stress properly), a spreadsheet provides all the advantages of hand calculations with a calculator and automated calculations with a dedicated statistics package. It is the best of both worlds, and that is why I use Excel to teach statistics when I am given the choice.

You may be interested to know how I happened to come to the conclusion that Excel is the best way to teach statistics. I was given the task of teaching a statistics class online, and found through that experience that a spreadsheet program was the only thing my remote students had in common. I was able to create spreadsheet models for the problems, and show the students how to do the calculations. Although that was relatively different for behavioral statistics, teachers of business statistics (of which I am also one) have used Excel for basic statistics for a long time.

More later.

What's Your Tool of Choice?

When I was a graduate student back in the dark ages, we had computers, but they were monstrosities that took up entire rooms. Some of the earliest statistical software I used was programs like BMDP and SAS. In those days, you wrote the commands for your statistical analyses in a syntax language, and typed that onto punch cards. You also had to learn JCL (job control language) to tell the computer and its operators where to get the data, which tapes to mount, and what to do with the output. You then input the JCL and program commands into the computer via a reader. After that, you went home or to class and waited. Eventually, your printout would tell you whether you had made a mistake in your code, and if you hadn't, your results would appear. It seemed like magic to me, as I had a professor who told me that you never really understand factor analysis until you do one by hand. No thanks!

About the time I finished my Ph.D., I invested in a good programmable calculator, a TI-55. It was actually pretty cool with its red LED display, and I used it to do t tests, ANOVAs, and correlation and regression analyses. I really liked that calculator and was sad when it died.

Today, of course, we have a variety of tools for statistical analysis. The TI-55 evolved into the TI-83 (and higher), and you have a lot of statistical prowess in your hand with a graphing calculator. SPSS and SAS morphed into desktop (and laptop) applications for Windows and Mac, and you can run more sophisticated analyses with your computer today than ever before. These programs are way easier to learn and use than they used to be.

But my favorite tool has to be Microsoft Excel. I don't use it for sophisticated or complex analyses, but I use it for almost all of my basic descriptive and inferential stats. A spreadsheet is the way I think about and visualize data, and running analyses in a spreadsheet helps me get a step closer to my data than I am with a statistics package. I can also do any number of what-if analyses, and see what happens to the results when I change the data. Another reason for using Excel is that you can save your formulas and functions in a spreadsheet template and use it over and over again for your statistical analyses. Look on the TwoPaces web site for various templates I have created for basic statistical analyses. I provide these free of charge for personal and educational use.

Larry

Take Statistics Online? You're Kidding, Right?

When I tell people I am a psychologist (which is true), they ask me to analyze them, or think I can somehow read their mind. So over the years, I have learn to answer the question, "So what do you do for a living?" with the simple (and also true) answer, "I teach statistics." That usually stops the conversation in its tracks, unless the person knows (or thinks he or she knows) something about statistics, or has had the misfortune of taking a statistics class from a bad professor, of which there are apparently plenty.

Very few of the students I teach have had positive experiences with courses in statistics, whether the classes were online or in the classroom. Attitudes toward statistics range from distaste to utter terror. Most students have the mistaken impression that statistics involves a lot of math. That is really not true unless you happen to be taking a theoretical statistics course. If you are taking basic statistics in an applied discipline, the only math you need is basic addition, subtraction, multiplication, division, exponentiation (squares and square roots), solving an equation for one unknown, and possibly logarithms. The hardest part of statistics is not the computations. Instead, the hardest part is understanding the logic of statistics.

My gift is the ability to explain complicated statistical terms and the logic of hypothesis testing in clear, understandable terms that are as simple as possible (but not oversimplified). I write about statistics, teach statistics at the undergraduate and graduate levels, and read about statistics for fun. I teach statistics online as well as in the classroom, and enjoy helping students and colleagues with their statistical problems.

In this blog, I will post information, answers, and thoughts about teaching and learning statistics. Please feel free to ask questions. In the absence of your questions, I will pose and answer questions based on my more than 30 years' experience as a statistics professor.