Nutrition 209: Mechanics

Nut209 is taught in two weekly 90 minute lectures and a total of two or three 60 to 90 minute structured computing labs.

Nut209

I give up! Uncle! I surrender! It seems there will never be a truly suitable text for Nutrition 209/309. Elemetary texts are too elementary. More advanced texts--even the good ones--are too imposing for someone getting his/her first real exposure to statistics and data analysis in these courses. And they are not cheap! And students end up relying on my notes posted at http://www.StatisticalPractice.com (as do many others [snapshot taken on August 13, 2007; snapshot taken on August 14, 2008; snapshot taken on September 11, 2009]).

There is no required textbook. For the last few years, the required text was Statistics at Square One by TDV Swinscow & MJ Campbell (BMJ Books) The 11-th edition, a $30 paperback published by the British Medical Journal, is readily available through Amazon. The 9-th edition, published in 1997, is available on-line for free and is not too different from the newer editions. It is a basic book.

In the past, I have used

• I considered Biostatistics by Wayne Daniel, published by John Wiley & Sons. I'd tried others (see below) but this was the one I keep coming back to. It's an introductory level text, quite readable, and presents estimation before hypothesis testing, but it does not contain many of the techniques discussed during Nut 309.
• Rosner B, Fundamentals of Biostatistics. The text "assumes only a knowledge of algebra." Rosner approaches the subject as a practitioner as well as a theoretician. However, he and I prefer to present topics in different sequences. I found that students were confused by having to jump around the text to follow topics in the order in which I presented them.
• Zar J. Biostatistics. Another nice book, but the level is probably too advanced for much of the class.
• Fisher & van Belle. Biostatistics. There is now a second edition (2004) with Heagerty & Lumley added as coauthors. The level is a step higher than Daniel or Rosner and relies a bit more on mathematical notation. It is comprehensive, though.
• Ramsey and Schafer, The Statistical Sleuth. Duxbury. In some ways it's my favorite, but it has two significant drawbacks. First, it's meant as a second text (not from the theoretician's viewpoint, but rather from the practitioner's viewpoint). It assumes the reader has been exposed to the basic techniques, perhaps in an introductory mathematical statistics course, and describes how to use them to analyze data. It's very similar in spirit to my online notes, but many students said that they wanted something at a more introductory level for their first text. The second drawback is that many of the examples may be drawn form subject areas that many might find uninteresting. I find that this, more than anything else, can dampen enthusiasm for learning. For example, one of the most important techniques you will learn is linear regression--fitting straight lines to data. Their first example, which is discussed in detail and used to motivate subsequent discussion, involves distance and recession velocities of extra-galactic nebulae. This is not the sort of thing FSNSP students deal with in their work or even in their private lives, so it is less likely to get students excited about the technique than an example from nutrition science or policy. Despite these problems, I've devoted many sentences to this text because I like the perspective from which it was written.
• Ott & Longnecker, An Introduction to Statistical Methods and Data Analysis. Duxbury. It has a wide breadth of coverage at an appropriate level. A large portion of its exercises involve interpreting computer output that the authors provide. Great for additional practice interpreting output.
• Kleinbaum, Kupper, Muller, & Nizam Applied Regression Analysis and Other Multivariable Methods, 3rd ed. Duxbury. This is an excellent text for learning about statistical modeling, but students found it difficult. The lessons it teaches in plain text are essential for anyone wishing to perform serious data analysis, but enough of the mathematics was beyond the scope of the course that students didn't enjoy it.

Other such textbooks can be found in the library under the call letters HA29. Everyone is encouraged to browse these books and use them as the mood strikes, especially to help master difficult material. If you find a particularly helpful book, please tell me about it.

Nut309

There is no required text for Nutrition 309. The most recent text used was Square Two, Second edition by Michael Campbell (Blackwell Publishing Limited, 2006, ISBN: 1405134903). That choice was made at the suggestion of students who found it helpful. I don't know of any text that covers all of the material at the right mathematical level. Also, I haven't found a text that I would care to follow chapter by chapter. I have my own view of the way the material should be presented, so adopting a particular text means either that I follow an order with which I'm uncomfortable or that students find themselves jumping around. I'd rather lecture and handout my own prepared notes.

There are a few other texts I think highly of, however. Students might consider them as a resource.

Computer-Aided Multivariate Analysis, Fourth Edition by Afifi, Clark, and May. It's interesting how my opinion of some texts and programs swings back and forth between editions. I liked the second edition very much and often referred researchers to specific chapters. I didn't care for the third edition. I like the fourth. I don't use it for a text because it does not cover analysis of variance. However, I thinking it might be effective to use it for what it does cover and continue to use my notes for analysis of variance.

Applied Regression Analysis and Multivariable Methods by Kleinbaum, Kupper, Muller, and Nizam. I tried using this as a text years ago for both semesters since it does an excellent job with simple linear regression, which is covered in 209a. It did not work well for the reasons stated earlier. Anyone who wishes to pursue a career analyzing data must eventually learn all this text has to offer. I tried once again in 2006-2007 after the class was divided leaving only the bench scientists and epidemiology students. The students would have be up for it, but I found that to use it effectively I would have to teach a much more mathematical course than the way I teach Nutrition 209/309. Since the focus is more on applications than underlying mathematics, I did not use it effectively.

Applied Statistics: Analysis of Variance and Regression by Mickey, Dunn, and Clark. I taught from the first edition, which was geared toward hand calculation, almost 30 years ago. This new edition strikes the right balance between discussion and theory. Earlier editions are quite different and will not serve as a substitute.

The structured labs will give you the basic statistical computing skills needed to get a statistical program package to do the calculations for you. Roughly once each week, I assign practice problems-- exercises to help you master the techniques presented in the lectures. Some students refer to them as practice problems for the exam. While I'd like to think that we are motivated by loftier ideals--the pursuit of knowledge, say--that is, in fact, what they are. More important, they will give you invaluable practice so that you will be able to apply the techniques to your own data. The only way to build skills is by doing lots of practice problems. You learn analysis by analyzing, just as you learn to drive by driving or to play a musical instrument by playing.

In previous years, a large batch of practice problems was assigned after each topic was completed in class. The problems were to be solved for in-class discussion the following week. My teaching assistants pointed out that for some students this meant making the same mistake repeatedly without ever mastering the technique. I now try break each set of problems in two. Typically, in any week you'll be working on some problems concerning the current topic (for the first time) and some problems concerning the previous topic (for the second time after having received feedback from your earlier attempts).

Each practice problem receives one of three scores: 0, 1, or 2. A 2 means you've got it right. A zero indicates a major flaw. A 1 is somewhere in between. That's as fine a distinction that I think it's reasonable to make. Many of the problems will require you to summarize your findings in a statement that might be printed in a report or journal article. If you do not word it properly, a 2 will not be awarded. Grading will be strict so that we do not reinforce the wrong behavior.

Nut209, Fall 2012 specific: There will be three exams

• the first on Wednesday, October 3,
• the second on Wednesday, November 7,
• and the third on Wednesday, December 17.

Both semesters: The exams are not designed to have students merely repeat what they did in class. There will be some of that, but much of the exam will involve synthesizing what was covered in lecture, lab, and practice problems, that is, putting it together in new and novel ways. You will be allowed to bring a sheet of notes and a calculator with you to the exams. You might wish to construct the sheet over the course of the semester rather than wait until just before the exam(s). At each exam, you will be allowed to bring the sheet prepared for that exam along with the sheets prepared for previous exams so that you be allowed four sheets for the Nut309 final exam. Hint: If you find you have any unused space, fill it with the solutions to judiciously chosen practice problems!

Grading is done with the instructor unaware of whose exam is being graded. At the exam, each student chooses a slip of paper at random with a number on it. The student's name goes on the slip of paper and only the number goes on the exam. Names are not linked to exams until after all exams are graded.

The one part of teaching that I can't stand is making up exams. It is hard to come up with good questions because I return exams to students, which means that there are 25+ years or exams floating around. I created a binder containing all of my old exams because some student groups are effective at compiling archives of old exams and I wanted everyone to be on the same footing. The binder gives everyone access to the same material. However...

The following rules will apply, mostly to preserve whatever little bit of my sanity still remains:

• Neither my teaching assistant nor I will discuss old exams. They're there for you to do what you want with them. They my prove useful in helping you study, but they may not if they cover material that is different from what we go over this year. It's up to you to determine what is what.
• I *may* reuse questions. If I do, I will modify them as much or as little as the mood strikes me.

Your letter grade will be based solely on the exams. The practice problems will be corrected and performance will be noted, but they will not figure into the letter grade, except in rare instances to judge borderline cases. I realize that most students put considerable work into these practice problems and many have asked why this effort is not directly reflected in the letter grade. There are three reasons for this.

• The first is that the effort is reflected, but indirectly. Those who work on the practice problems will be in better shape going into the exam than those who do not.
• Second, practice problems are just that--practice! They give students it opportunity to experiment and get it wrong without having it affect their letter grades. If they were a determinant of the letter grade, it would be wrong to call them "practice".
• The third reason is that one year I experimented with letting the practice problems count toward the final grade. Not only did students feel pressure to never make a mistake, but also the entire learning process was perverted. In the eyes of some students, the purpose of the practice problems was no longer to build skills but to get the best score possible, even if that meant vigorously arguing untenable positions. That is, it became more important to argue for full points than to understand why they weren't awarded. Nothing I tried could stop it, including offering to throw out the three lowest grades. This behavior did not characterize all students, but it was true of enough of them to make me decide never to do it again.

Exams: Your exams are ALWAYS to be entirely your own work. You will NEVER be permitted to work with another student. You will NEVER be permitted to have contact with another student about the content of an exam while it is in progress. If you ever see instructions that suggest otherwise, you are misreading them or I have mistyped them.

Practice problems: With practice problems, I present you with a connundrum: You are encouraged to work together, informally or in formal study groups as you choose, but I would like you to turn in your own work.

Building skills is hard work. We all need all the support we can get. It's easy to say what is inappropriate on exams--any contact with anyone, including your classmates, is inappropriate and constitutes of academic dishonesty. This is also applies to the take home portion of the exam. On the other hand, with practice problems I expect you to serve as resources for each other. You're probably doing it right if

• you are not merely copying but are writing out your answers yourselves, and
• you could do it all without assistance if you were starting over from scratch.

I would appreciate any insights you might have into how to achieve these seemingly inconsistent goals of supporting each other while doing your own work.

Study groups are a good thing. Form them! I don't have a formal mechnanism for getting people together. They seem to do it on their own...if they think of it. Think of it!