One of the bigger problems with the statistics courses (Bus., Psych, or Math) that we accept/require for undergrads is that they have so very little to do with anything relevant. Out of frustration, I created a course for majors that begins with data collection and ends with a presentation and covers the steps along the way, and in so doing I can easily incorporate the statistics needed. For what, >90% of all questions?, we need Chi2, t-test (1-way AOV), correlation, and/or regression. Very simple stuff. We get in trouble when we go deeper than that without the training or experience. We do not need less training in statistics, we do need more appropriate instruction.
For example, to my recollection, a most valuable piece of advice from my diss. adviser was to stop focusing on the means and look at the variation, particularly on graphs. In other words, learning to interpret statistics has to do with knowing what is important in a statistical sense. For my beginning students, the ABCs are randomization, replication, and control, with considerable focus on the importance and influence of sample size. On the other hand, the course isn't really about statistics, it's about experimental design and then that leads into the appropriate tests for the particular questions. My vote? Keep the statistics, but be careful letting others teach it. Cheers, Andy Andrew R. Dyer Asst. Professor of Ecology Dept. of Biology & Geology University of South Carolina Aiken 471 University Parkway Aiken, SC 29801 Vox 803-641-3443 Fax 803-641-3251 [EMAIL PROTECTED] -----Original Message----- From: Ecological Society of America: grants, jobs, news [mailto:[EMAIL PROTECTED] On Behalf Of Bill Silvert Sent: Thursday, November 03, 2005 12:35 AM To: ECOLOG-L@LISTSERV.UMD.EDU Subject: Re: curriculum question I didn't expect much agreement with my posting, and I'll just comment on two points that Roper raises, interspersed with his posting below: ----- Original Message ----- From: "James J. Roper, Consultor - Tradutor" <[EMAIL PROTECTED]> To: <ECOLOG-L@LISTSERV.UMD.EDU> Sent: Wednesday, November 02, 2005 9:28 PM Subject: Re: curriculum question > Every really good paper published in Ecology and > many other ecological journals required the statistics that was included > to make their point. True, but that doesn't justify the statistics. I've taken a cynical approach to this, on occasion when my colleagues have asked if I can suggest statistical methods to analyse the data I usually find that the conclusions are obvious just from looking at the data - so I discuss the conclusions with them, and then once we know the result they find it easy to come up with enough statistical filler to get the paper past the referees. > Bill goes on to contradict himself when he says: > > "In terrestrial work where > sampling tends to be easier and one can lay out quadrats on foot, etc., > statistical methods can be very useful." It's not really a contradiction. I pointed out that there are exceptions, and my criticism of the use of statistics was not categorical. Moderation in debate is not always a vice, no matter what Barry Goldwater said. Bill Silvert
