Dear All, I appreciate all of the input I received about biostatistics texts. I've pasted the original post plus a compilation of responses below (be forewarned, it's about 8 pages of text!). I found the two posts from Howie Newfeld and the response to them (first 3 responses below) to be particularly helpful.
I found people to be relatively split among several camps: 1. the tried-and-true old standards (Sokal and Rohlf or Zar) 2. newer books with a more ecological bent (Gotelli and Ellison, Quinn and Keough, Scheiner and Gurevitch) 3. more user-friendly, readable intro books (Havel and Hampton, Glover and Mitchell). 4. a few particularly liked books that use R or S+ (Crawley, Dalgaard, Venables and Ripley). Gotelli and Ellison's Primer of Ecological Statistics, in particular, got rave reviews from a number of people. Personally, I'm thinking about using either Hampton and Havel (Introductory Biological Statistics) or Glover and Mitchell (An Introduction to Biostatistics) for my introductory class. For my followup grad class, I considering Quinn and Keough (Experimental Design and Data Analysis for Biologists) and possibly supplementing it with Gotelli and Ellison and/or Scheiner and Gurevitch (Design and Analysis of Ecological Experiments). An article comparing different biostats texts would be a nice project for someone to undertake for the ESA bulletin or some other venue!!! Any other parting shots about biostats texts? Mark D. Dixon Assistant Professor Department of Biology University of South Dakota Vermillion, SD 57069 Phone: (605) 677-6567 Fax: (605) 677-6557 Email: [EMAIL PROTECTED] ************************************************************************ **** Sent: Wednesday, October 18, 2006 9:21 AM To: [email protected] Subject: Biostatistics texts Does anyone have recommendations for a text for introductory biostatistics? The class is junior/senior level course with mostly students with an ecological/ environmental bent, although there may be some pre-meds as well. From my discussions with others, Zar seems to be the top choice, but I was wondering about other possible contenders (as well as any feedback folks have on Zar). Any input would be most appreciated. ************************************************************************ **** Mark - I teach an Intro Stats course to incoming graduate students, but the content could easily be grasped by juniors/seniors (I assume no a priori knowledge of stats on the part of our students). For many years, I used Zar. Zar is good because it has many examples worked out for the students. However, the drawback is that it contains a lot of extraneous material which I never use, and it's expensive. Thus, this year I began an experiment by switching to a relatively new, and very inexpensive book ($25, paperback). It contained nearly all the material I used to cover in my one semester course, but without all the extra content. Further, the graphics are fairly well done, and the writing is lucid and clear. There are also plenty of work problems, and a CD that comes with the book. My students so far say they really like the book. Here is the citation: Introductory Biological Statistics, 2nd ed., by R.E. Hampton and J.E. Havel. 2006. Waveland Press Inc., Long Grove, IL. ISBN 1-57766-380-2. The press' phone number is: 847-634-0081. Waveland Press Inc., 4180 IL Route 83, Suite 101, Long Grove, IL 60047-9580. www.waveland.com The one drawback is the paucity of statistical tables at the end, and the fact that Havel (the corresponding author, as the lead author has died) uses a Z table that gives the P for the area under the normal curve between a value of 0 and the calculated value of Z, rather than the P for the proportion of the curve in the tail. Havel has told me that he may change it to the more usual table of values in a future edition. He does provide a set of URLs where students can get more detailed statistical tables. But for the most part, everything they need is in this book. ************************************************************************ ******** Dear All - the thread about which statistical text to use is interesting to read, and reflects, I think, the depth and breadth of statistical sophistication among us ecologists. Those of us with a smattering to moderate amount would probably prefer a more introductory book for our incoming graduate students, while those with a lot of training in statistics would prefer a higher powered book. For years I used Zar - it does have depth and breadth, and plenty of examples worked out. It's a great reference book too. Others seem to prefer Quinn and Keough. And I'm sure there are yet others who have their own favorite texts. Based on my sampling, I think the only true conclusion is that medical statisticians write the absolutely worst texts and ecologists the best ones! However, this year I switched texts to one by Hampton and Havel (H&H) because I realized that in my intro biometrics course (that's what we call it here historically) I never got to at least half of the chapters in Zar, not to mention that the Zar book is expensive. Much the same material I covered in past years is included in the H&H book. But because its paperback, it has a greatly reduced price (~$25). It's also geared for those students with essentially no a priori background in statistics, and my students like it so far. I found the Quinn and Keough book way too advanced for the introductory student. From a pedagogical point of view, I thought it was poorly developed. It had its moments, but as a teaching tool for students just starting out, it would have been way to much for my students. It too contains much more material than I could ever cover in my intro course. Not wanting my students to have too much disposable income, I supplement the H&H text with Gotelli and Ellison's new Primer on Ecological Statistics, because I like their philosophy and approach to statistics. However, they provide no work study problems, but again, and as Aaron Ellison has told me, that was not their goal. But their discussions of why we do statistics, the history involved, and their section on experimental design are all highly readable, so I assign this as an auxiliary text for them. I especially like their discussion of what Bayesian statistics are, and how they can be used. That is not something most of even mention in intro courses. Finally, we conduct a SAS lab each week. Yes, I'm one of the dodos of the statistical world that still finds SAS programming useful, and so I inflict this on my students (If I had to do it.....! - no that's not the reason!). For this we use Cody and Smith's "Applied Statistics and the SAS Programming Language". I'm sure there are those who find this type of training anachronistic, but simply using point and click programs often leads to errors in experimental design and then analysis, which are less likely if you are writing the programming yourself. By the end of their first semester, they can move on to point and click programs, so they end up with several skills here. The main problem we have at Appalachian State is a follow-up experimental design course for our graduate students. I would be most happy if someone who is teaching a second semester course in this area would send me their syllabus. We want to set up such a course here, and I would appreciate feedback as to what we should include in such a course. Thanks for listening to this long ramble. Howie Neufeld Hello Howie - (I am one of the few poor souls who actually enjoys reading "long rambles" about imparting good statistical know-how to future scientists :) I can definitely see your point about the difficulties associated with implementing the Q&K text in introductory courses. As much as I do like it, I have yet to adopt it in my courses. However, I do like the fact that they avoid a 'cookbook' approach, and present quantitative methods within the broader context of conducting science (and thinking about what your are doing!). This is also a strength of Gotelli and Ellison's book, which I also like very much. In fact - although I only require one textbook - I try to emphasize to my students the value of starting their own library, and encourage them to at least take a look at all of the texts which have previously been mentioned. For anyone wanting to pursue a career in science - good resources (and especially stats resources :) are a worthy investment! I will check out the H&H book you mentioned. And I wholeheartedly agree with your aversion to teaching "point and click" approaches to statistical computing!! In addition to reducing error, forcing the use of explicit code creates a reproducable and well-documented history of past analysis (and data management). I have incorporated R into my introductory course, and have had very good success. Once students get over the initial learning curve, they quickly learn to appreciate the power and flexibility of knowing a statistical programming tool. "Introductory Statistics with R" by Peter Dalgaard is a great supplementary text for this purpose. Count me in the camp that appreciates Zar as an introductory text. I have not taught a stats class, but Zar is the text assigned for a stats class I had many years back. I remember it being good at introducing the concepts and I continue to use it as a reference today. Another book that i like that hasn't been mentioned yet is "The Ecological Detective" by Hilborn and Mangel. While not a good introductory text, it might be good as a supplemental book for discussion of use of statistics in ecological research. And since Stephen mentioned R... I have learned most of what I know about R from Dalgaard's Intro text in concert with the more complete "Modern Applied Statistcs with S" by Venables and Ripley. I used Sokal and Rohlf, but again not by choice (I was bound by tradition and other considerations). I found the students relied on the boxes, which are very well done, but were mystified by the text. There are also a number of errors in their motivations. I think Zar is the better text, and much more accurate, but it's also a little more advanced in its presentation. I used it at Midwestern University, and the students there found it to be too difficult. (Of course, the students there are generally not nearly as motivated towards research as most BIO 415 students are.) I've also used another text on Medical Statistics, the title and authors of which I can't remember largely because it wasn't very good. It, like many of the other books I reviewed, was very mechanical and treated everything algorithmically. It gave the students no foundation on which to understand what was going on. The Primer of Ecological Statistics by Gotelli and Ellison is a pretty good text that is oriented towards students with an ecological/environmental bent. I find Quinn & Keogh to be an excellent text for biologists/ecologists. They also provide online datasets. Quinn, G.P. & Keough, M.J. (2002) Experimental Design and Data Analysis for Biologists Cambridge University Press, Cambridge. http://www.amazon.com/Experimental-Design-Data-Analysis-Biologists/dp/05 2100 9766 Given your brief, may I suggest "Practical Statistics for Field Biology" by Fowler et al. John Wiley & Sons. It is a great baseline book for anyone with basic statistical knowledge, explains the theory very well and uses good examples. Its also pretty cheap as far as reference texts go and I would certainly recommend it. Although I find Zar very good, I would not call it basic and therefore it depends on what you are looking for. In my opinion, Zar is far better than Sokal and Rohlf (in particular). A few years ago I did an informal survey on EvolDir of this question. Of the 40 respondants, about 15 favored Zar and 15 favored Sokal & Rohlf. The remainder were diverse in their tastes, with no more than two suggesting any other book. The major problem with both of these leading contenders is that they are old school statistics. Both are in revision, but I do not know when the new versions will be out. Zar is more of a cookbook, whereas S&R provided lengthy discussions of the models and what they mean. Last year I used a text "in development" by Mike Whitlock and Dolph Schluter that was very good, but short on linear models. The final version may be somewhat different. I also have looked at Modern Statistics for the Life Sciences by Alan Grafen and Rosie Hails that takes a more model-based approach than any of the above; I might use that the next time I teach. If I had to choose today, I would stick with Sokal & Rohlf. Students hate it because of the lengthy explanations, but I find several chapters invaluable in laying out the assumptions that underly the models and in developing statistical thinking in the students. You have to assign limited pages to keep the students happy, but it is a great, if somewhat dated, reference. Zar is fine if you just want students to analyze data, but I find little help in the more difficult task of understanding the statistics. Whitlock & Schluter have a great knack for explaining how to find the biology in the stats, but it really is an elementary statistics text at the last point I saw it. If you are looking for a computer oriented text, Michael Crawley's Statistics: An Introduction using R is superb. Very modern, very R-centric. You will learn cool stuff yourself from this book. I use this as a supplement for S&R and the students love it. I think that any text today should leverage on computational tools; nobody does ANOVA by hand, yet some of us spend a lot of time teaching stats that way. I have students compute their first sums of squares "by hand" in R, and build their own ANOVA table, plus do a couple of proofs with pen and paper, but mostly we use the real tools that every one of us uses professionally. R is great because a) it is free, b) it is modern, c) it allows easy simulation, bootstraps, etc. to illustrate important statistical concepts, and d) it makes beautiful graphics. The first few weeks are hard, but when they complete my course they have a marketable skill that will serve them no matter how far they go in biology. I have found 'A Primer of Ecological Statistics" by Gotelli and Ellison (2004; ISBN 0878932690) to be of great help. And even though it has the word 'ecological' in the title, it is well-written and general enough to be used by many classes. And at less than $40, your students won't complain about the price. Try An Introduction to Biostatistics by T. Glover and K. Mitchell published in paperback by Waveland Press, ISBN 1-57766-459-2. It includes a CD with an instructors manual and answers to the problems. It's great for biologists I've been teaching Biostats for a couple of years and used Zar up until this semester. Zar is a great reference book. But for the course that you described (similar to our BIO 3105) the two problems are: 1) students don't read it. The first draft predates many (most?) stats software and the text spends a lot of ink on detailed calculations that we really don't cover. Also - you would only use 20-30% of the book. 2) It is expesnsive. This semester I've switched to Havel and Hampton. It's ok, inexpensive, and we use it cover-to-cover. There are a bunch of new stats books out there and I suspect they are all vying to be the next Zar. Unfortunately, they all suffer from "textbook-itis" - they have expensive hardbound versions with non-recyclable paper and lots of irrelevant photographs. The H&H book is a good, inexpensive alternative. Good luck. I would suggest Gotelli and Ellison's "A Primer of Ecological Statistics" (be sure to get the second printing, see http://harvardforest.fas.harvard.edu/personnel/web/aellison/publications /primer/Errata/errata.html). Our library here also recently got "Stats without Math" (http://www.sinauer.com/detail.php?id=5061), which I skimmed through. It might be a great low-level intro, although I have never tried teaching from that approach. I was trained in the dense true stats books and they're just tough to make accessible, no matter how you do it. There is a fairly new book called "A Primer of Ecological Statistics" by Gotelli and Ellison. It is fairly popular among ecologists here at UofL. No matter which statistics book you use, you might also try "Statistics without Math" by Magnusson and Mourao as a supplement. It offers explanations of statistical tests and concepts without getting bogged down in formulas, which has certainly helped me understand statistics. It is also fairly short (136 pgs). I saw your list-serv posting asking about biostatistics texts. I had Zar as a MS student and found it next to impossible to read. I wasted money on it, never used it and sold it back as soon as possible. To my mind there is nothing better than "Biometry" by Sokal and Rohlf. It's clear, readable, uses relevant examples and leaves the student feeling capable of handling the subject. In contrast, Zar left me with a sense of hopelessness. After reading Sokal and Rohlf, however, statistics felt do-able. Feeling competent to handle a subject should never be underestimated. I can say that I have also used Steel and Torrie's text and found that to be more approachable than Zar as well. For your non-majors students, stay away from Zar. Go with Sokal and Rohlf. I used Zar as a junior taking my first course in statistics, which was taught by a member of the biology department at Tufts, Sara Lewis. I found it very accessible and a good supplement to her lectures. My math background went through the third semester of calculus; I assume most of your students have had at least a semester or two of calculus and are roughly at the same level. The course I took included problem sets, which were essential in terms of helping me understand the material. Most of the problem sets were to be done by hand, but we had some exposure to statistical software as well. I personally pretty much like the following: Quinn G.P., and Keough M.J., 2002. Experimental design and data analysis for biologists. Cambridge University Press. A fairly accessible text with real data examples but which also goes into details. May be not as much "introductory" as Zar though... there is also the Biometry by Sokal and Rohlf (2000, W.H. Freeman and Company) which remains an excellent reference. Even less "introductory" and somewhat frightening at first but actually very clear and useful. I've also heard good comments about "Intuitive Biostatistics" by Motulsky and Motulsky, Oxford Univ. Press but I personally never had it in hands... the department of biology at the university of denver uses Biometry by Sokal and Rohlf and personally it is quite limiting and confusing to even graduate level students. it is a great desk reference, but only if you are already familiar with common statistical theories. the books that i strongly recommend to all incoming graduate students for desk references follow: A Primer of Ecological Statistics. 2004. Gotelli, N.J. and A.M. Ellison. and Research Design and Statistical Analysis. 2002. Myers, J.L. and A.D. Well. each book takes the time to stress the importance of research design and how that adds to the strength of your statistical model. I recommend Sokal & Rohlf - Biometry. It looks a bit intimidating at first but it was used as the main text when I did Biostatistics at both BSc and MSc levels and I continued to use it during my PhD. Zar is often the preferred text and seems to be the text most often cited by ecologists. Personally, I find Zar to be denser than my students are prepared for. Our students here are generally math-phobic (ironic for science majors) and I fear that a dense stats book would actually discourage them from using the book. So I've gone in the opposite direction thus far and used an extremely basic text by Hampton and Havel (Introductory Biological Statistics, Waveland Press). I supplement the text as needed for a few topics not included but find that the text presents stats in a very understandable, straightforward manner that my students can grasp and appreciate. In my syllabus, I include a list of additional texts (list at end of message) that may be of use and strongly encourage graduate students or aspiring graduate students to purchase Zar for their personal library. Admittedly, Hampton & Havel has its limitations and I have been considering alternatives including: - Grover & Mitchell. An Introduction to Biostatistics (Waveland). This seems to be a good compromise between accessibility and detail (and $$), and I will give this text serious consideration before my next course. - Ott & Longnecker. A First Course in Statistical Methods (Thompson). This is the text that I would use if my students were less challenged quantitatively - excellent presentation, relevant (& variety of) examples, great context/descriptions. The same authors also have another text: Statistical Methods & Data Analysis. In short, I think that it depends in part on the students' abilities and your overall objective for the course. Best of luck. Zar. 1999. Biostatistical Analysis (4th ed). Prentice Hall Sokal & Rohlf. 1995. Biometry. W.H. Freeman Scheiner & Gurevitch. 2001. Design and Analysis of Ecological Experiments. Oxford University Gotelli & Ellison. 2004. A Primer of Ecological Statistics. Sinauer Associates, Inc. Underwood. 1997. Experiments in Ecology. Cambridge University Press. I recently picked up "A primer to ecological statistics" by Gotelli and Ellison and was really pleased with the sections on statistics. Gotelli also wrote "A primer of ecology" which I also highly recommend. Gotelli's writing is much clearer than other texts with out loosing the meaning. Textbooks like Zar can be a bit dry and overwhelming, especially for undergraduates. Since your class with have many people interested in ecological applications I think this book would serve you much better. Crawley 2002 Statistical Computing. It uses R (which is open source and great) or Splus. The book explains a lot of stats, unlike Zar which is pretty cookbook. And Crawley is an ecologist. Zar is good. You should consider Biostatistics by Sokal and Rohlf. This is a maserpiece in my opinion. Has almost everything that any biology student might need. But the key is to get the first few chapters straight. If you want your students to think of statistics as "math magic" by all means use ZAR. It is full of mid 20th century pre-computer statistical thinking, and is poorly written. If you want your students to be able to think about how the math relates to the biology and if you want them to understand what the statistics are doing, get A Primer of Ecological Statistics by Gotelli and Ellison It is clear, concise and explores the mathematical concepts rather than showing copious ammounts of simple mathematics done poorly with overly detailed notation. Generally speaking, computers can do the simple repitious calculations that seem to clog up many statistics texts. It is more important to understand the purpose of the math and how it relates to the ecology. Gotelli does this well. Having said that, Gotelli does not provide practice problems and you will have to choose what subjects to cover. on the plus side, it is relatively inexpensive and is an excellent reference to refer back to once you have read it. PS I am currently teaching out of ZAR (not my choice) and find myself supplementing my lectures with details from Gotelli. Hi Mark - I have used Zar in an introductory course for several years. It is very good text at balancing breadth and depth. That said, if I was starting over - especially for students with the ecological/field biology focus you mentioned - I would seriously consider the Quinn and Keough text that was suggested. This is a fantastic book! Take a look at both and see which fits the topics you wish to cover in the manner in which you would like to present them. As an Ecology grad student, who took two killer biostats courses, I would say that there is no better book than Quinn and Keough. Zar is outdated and less clear. Moore and McCabe is ok, but less bio related. Now that I'm further along, I reference Neter et al. as well (Applied Linear Statistical Models) but it would be hell for an entry course. I would go with Quinn and Keough. I haven't taught an intro biostats course, but I've taught a grad level Biometry course now at two universities. At a middle sized state school with a master's program I used Sokal and Rohlf's Biometry text with some other readings. I thought it was a good text. Some of the students had problems with it, thinking it was too meaty and somewhat difficult to read. Still, the students who lasted in the course seemed to come around to it over time. I recently began teaching Biometry here, at University of Arkansas, and I used Quinn and Keough with some supplementation with other readings. The Quinn and Keough seemed to be much more highly rated by the students. Personally, I think it's an outstanding text and the style and organization mirror my teaching extremely well. Therefore, I'll probably stick with Quinn and Keough for the foreseeable future. We use Zar for Biometry at ASU. I took the course and now TA it with John Sabo and I think using Zar has been successful. We supplement with readings from Scheiner & Gurevitch and the first chapter of Burnham & Anderson but our course is mostly grad students with a few undergrads. We also read papers from the primary literature each week that exemplify each test or element of experimental design covered in lecture/lab. I have also skimmed Ellison & Gotelli's text that introduces basic statistics with a Bayesian vent. I think it aims to be too controversial although there may be sections worth assigning to give students a Bayesian perspective.
