I'm not sure I understand Manuel's distinction between statistical
hypootheses and scientific hypothesis. Is not the former supposed in some
way to mathematically embody/parameterize the latter?
But in any case, it seems to me that it is often hard to rigorously
formulate a null hypothesis and a corresponding working hypothesis. Suppose
you hear an account where someone had a feeling of foreboding about his
mother, only to discover later that just when he was having that feeling,
his mother, thousands of miles away, had suddenly died. When people tell
stories like this, it's often followed with a challenge, like "you can't
tell me that's just a coincidence!"
Well, I'd like to say it is a coincidence, but how could you test
it? What is the expected number of times you should have a feeling of
foreboding about your mother and she DOESN'T die? What is the expected
number of times mothers should die without their sons/daughters having
feelings of foreboding? How close to the actual time of death does the
feeling of foreboding have to be before we can count it? How creepy does a
feeling have to be before it reaches the threshold of genuine foreboding?
Now, this doesn't sound very ecological, but I'll bet readers of this
listserv can come up with examples from biology that approach this level of
nebulosity. Here's my stab at it: How K-selected must an organism be before
we say it is K-selected (or r-selected). How many factors in an environment
must conduce to K-selection before we say it is a K-selecting environment?
How many species in that environment must bear the earmarks of K-selection
before we accept the hypothesis that it truly is a K-selecting environment?
What about all the species in that environment that don't appear to be
K-selected?
I realize, of course, that different organism may be responding to
different factors in the environment, and that we can get around some of
these problems by defining a hypotheses sufficiently narrowly. However, the
more narrowly we define the hypothesis, the less it tells us about nature
because it is less generalizable, and I suppose that most researchers would
like to come up with insights that are generalizable.
I don't know if this relates to some of the problems that prompted
Jane's query, but I'd love to see your thoughts on the matter.
Martin M. Meiss
2011/2/28 Shermin ds <shermi...@gmail.com>
> I like Manuel's response.
>
> To answer Jane's other questions:
> 1. Does it help you do better science?
> It can, but not necessarily. See below.
>
> Is it crowding out other approaches?
> I'd like to hear more about this - what other systematic approaches are
> there? For example, anecdotal observations are generally discouraged, but
> sometimes anecdotal observations are valuable and should be a) reported and
> b) inspire further observation and/or experiment. E.g. observations of
> tool-use in animals in the wild are great example of spontaneous events
> that
> one can never set out to observe systematically (except in controlled lab
> settings) but are nonetheless highly informative.
>
> I also wonder about replication - the larger or longer the scale (e.g.
> ecosystem, biome/longitudinal studies) the harder it is to replicate. This
> gets at Manuel's distinction about statistical vs. scientific hypotheses.
> You might have a hypothesis about a process but observe outcomes that are
> inherently difficult to attach a p-value to or find multiple examples of.
> Thoughts on that?
>
> Finally there's the issue of taxonomic poverty. Hypotheses about clades
> with few species are more difficult to test than those with a greater
> number
> of species. A problem if you're interested in the species-poor clade for
> other reasons. I.e. there is a trend towards choosing your species/system
> of
> study based on your questions of interest, and lately I've heard many talks
> that begin "We chose to study species X because it is an excellent model
> for
> testing Y..." What if you simply want to know about species X for no other
> reason than that you want to know about it?
>
> 2. Have you ever had a grant proposal or publication declined because
> of an absent or unclear hypothesis?
>
> Yes, and I'm wondering about this trend in the stated aims of some journals
> as well.
>
> --
> Shermin de Silva, Ph.D
> http://elephantresearch.net/fieldnotes
> http://www.sas.upenn.edu/~sdesilva
>
>
>
> On Mon, Feb 28, 2011 at 10:49 AM, Manuel Spínola <mspinol...@gmail.com
> >wrote:
>
> > Dear Jane,
> >
> > That is a topic that have interested me for a long time. I teach
> something
> > of this in my classes to master students in wildlife management and
> > conservation here in Costa Rica. I know this is a controversial issue.
> >
> > First I recommend these 3 books:
> >
> > Scientific Method for Ecological Research. E. David Ford.
> >
> > Method in Ecology: Strategies for Conservation. Kristin S.
> > Shrader-Frechette and Earl D. McCoy
> >
> > A Primer on Natural Resource Science. Fred S. Guthery
> >
> >
> > Is necessary to distinguish between statistical and scientific
> hypothesis.
> > Statistical hypotheses is about patterns, scientific hypotheses are about
> > process (they are based on "why" or "how").
> >
> > My experience on this topic tells me that most ecologists do not know the
> > difference between the 2 kind of hypothesis.
> >
> > Like you probably experienced, reviewers like to see hypothesis driven
> > research on the proposal that you submit but most of the time they do not
> > know what a true scientific hypothesis is.
> >
> > Most research in ecology is not hypothesis driven, even when would like
> to
> > see that. Read any paper in ecological journals and see how many of them
> > are truly hypothesis driven.
> >
> > Hypothesis driven research are not always possible and in many instances
> is
> > not necessary to have scientific hypothesis, all depend on the context.
> > Most of the time we are interested in parameter estimation on how much a
> > factor or covariable influence a parameter of interest. Besides, If you
> are
> > going to do hypothesis driven research you need to work with multiple
> > hypothesis (Chamberlin).
> >
> > Falsification is the contribution of Karl Popper to the
> > Hypothetic-Deductive method. It has nothing to do with statistics or
> > statistical hypothesis.
> >
> > The hypothetic-deductive method has been considered as "the scientific
> > method", however not many people know how it works. The
> > hypothetic-deductive method is inductive and not deductive like the
> > namesuggest.
> >
> > There is no a superior approach to obtain scientific knowledge.
> >
> > There are much more on this topic but I would like to see other opinions.
> >
> > Best,
> >
> > Manuel Spínola
> >
> >
> >
> > On 27/02/2011 11:44 p.m., Jane Shevtsov wrote:
> >
> >> Fellow Ecologgers,
> >>
> >> Lately, I've been thinking a lot about the role of hypothesis testing
> >> (both the statistical and falsificationist varieties) in biology in
> >> general and ecology in particular. Before saying anything, I want to
> >> ask the forum a few questions.
> >> 1. What do you think of the current emphasis on hypothesis-driven
> >> research? Does it help you do better science? Is it crowding out other
> >> approaches?
> >> 2. Have you ever had a grant proposal or publication declined because
> >> of an absent or unclear hypothesis?
> >> 3. Have you ever recommended that someone else's grant proposal or
> >> publication be declined for that reason? Was it the main reason?
> >>
> >> I look forward to hearing what people have to say.
> >>
> >> Jane Shevtsov
> >>
> >>
> >
> > --
> > *Manuel Spínola, Ph.D.*
> > Instituto Internacional en Conservación y Manejo de Vida Silvestre
> > Universidad Nacional
> > Apartado 1350-3000
> > Heredia
> > COSTA RICA
> > mspin...@una.ac.cr
> > mspinol...@gmail.com
> > Teléfono: (506) 2277-3598
> > Fax: (506) 2237-7036
> > Personal website: Lobito de río <
> > https://sites.google.com/site/lobitoderio/>
> > Institutional website: ICOMVIS <http://www.icomvis.una.ac.cr/>
> >
>
