At the risk of repeating another opinion, I could not think that Bill
could be more wrong!
A skill in any field comes from knowing how to use the tools. I would
take Bill to task to "prove" what he says about little value from
statistical analysis. Every really good paper published in Ecology and
many other ecological journals required the statistics that was included
to make their point.
Sure, there are many bad examples of statistics, and I would bet that
Bill is also wrong about most ecologists being able to spout ANOVA and
t-tests in their sleep! And, 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."
Also, where he says that
The literature of the field is full of schemes for stratified random sampling
and negative binomial distributions, but virtually no real ecology.
This problem is NOT due to statistics, but rather the particular
situation. And, could it be that those in this field don't know their
tools or their theory?
And, if we were to wait for good teachers, many other things would be
forgotten! What we need are good learners and self-learners.
Statistics, to be useful, must be understood and anybody who uses tools
well, understands those tools. Otherwise they should hire out.
Jim
Bill Silvert wrote:
>At the risk of repeating myself I feel compelled to respond to Ryan Walker's
>post. Unless the teaching of statistics can be totally changed, I would
>argue for less statistics, not more. As I have pointed out before, most
>ecologists can spout ANOVA and t-tests in their sleep, but almost none can
>do something as basic as adding two numbers (remember my earlier post about
>adding 100+-3 to 200+-4?). Most statistics courses deal exclusively with
>linear models to the extent that the majority of books I have surveyed hew
>to the old line that transformations are for the purpose of linearising data
>(they should be used to normalise variances).
>
>Over all I have seen little of value come out of statistical analyses, which
>usually just confirm the obvious, but I have some incredibly stupid
>conclusions drawn from incorrect use of statistics. In balance I think that
>the value of statistics is not significantly greater than zero, if indeed it
>is positive at all.
>
>Of course this can vary with the subfield. In terrestrial work where
>sampling tends to be easier and one can lay out quadrats on foot, etc.,
>statistical methods can be very useful. The use of statistical models in the
>design of agricultural experiments is clearly essential for example. But in
>areas where data are collected in a more opportunistic way the use of
>statistics is often a diversion rather than a help. In aquatic ecology, and
>especially biological oceanography, statistics can be a real nuisance - if
>anyone ever captured the Loch Ness monster they couldn't publish the news
>because one is not statistically significant!
>
>For a particular example of what I mean, look at fisheries oceanography. The
>literature of the field is full of schemes for stratified random sampling
>and negative binomial distributions, but virtually no real ecology.
>Basically the statistics has edged out the ecology, and it is too hard to do
>both.
>
>Bill Silvert
>
>
>- Original Message -
>From: "Walker, Ryan" <[EMAIL PROTECTED]>
>To:
>Sent: Tuesday, November 01, 2005 6:18 PM
>Subject: Re: curriculum question
>
>
>
>
>>Having come from a good undergraduate program (University of Wisconsin -
>>Stevens Point) and working on a graduate degree at a university with a
>>somewhat lacking undergraduate program (Texas Tech University), I have
>>seen both sides of the coin. Regardless of the focus of the program
>>(Ecology, Wildlife Management, etc.), there is a general need for more
>>statistics and experimental design. My undergraduate program was more
>>focused on management and techniques of wildlife ecology and its limited
>>statistical requirements are still more than other programs. Focusing on
>>statistics that may be useful for students to know, such as
>>non-parametrics and multi-variate analyses. I realize that students may
>>have difficulty grasping some of these more complicated topics, but I feel
>>it is necessary to expose students to this material. A simple knowledge
>>of the tools that are available for research would be extremely helpful.
>>
>>
>
>
>
>
>
--
Atenciosamente,
James
=
Consulto ECONCIÊNCIA
Ecologia, Conservação, Ciência e Consciência!
Consulting, specializing in Conservation
Research Methods, Analysis and Translations.
http://www.montanhaviva.org/ecosci/
=
James J. Roper, Ph.D.
Caixa Postal 19034
81531-990 Curitiba, Paraná, Brasil
-
E-mail: [EMAIL PRO