[Aqualist] silt/sand boundary

Tue May 27 13:13:37 EST 2003

    There are some possible problems with the statement by Erlich
(not Erlich's fault!):

It is necessary to separate statistics into two categories:
useless and useful.  The rise of personal computers means
that people have generally forgotten the important fundamentals
that lie at the heart of experiments and statistics
(eg R. A. Fisher's work in the 1950's); there are too many papers where
a computer is used to do something that is useless.
The point of statistics has been forgotten: it provides
a way to make inferences about populations via samples,
and I haven't seen many papers where people have sampled the
entire population of sediments in a river bed.
The difficulty with sediment samples is that conventional 
statistics does not handle samples which are made up
of % compositions.  Thus it is not possible to use conventional
statistics on a sediment composition to answer the question:
'When does the 95% confidence intervals for my sediment samples
fall within 1% of the true population?' (note that I haven't
had to use words such as mean or median here!).  The techniques examined in
the weltje paper appear to be the only techniques which can
actually be used to answer that question for a % composition.

A second point here is that other disciplines such a climatology,
ecology have profited from useful statistical analysis
(eg see von Storch and Zwiers.  1999.  Statistical analysis in climate Research)
These disciplines have also suffered from useless statistics!

Erlich's point is valid because it is about useless statistics,
but it is not about useful statistics.  It IS time that we started
teaching students useful statistics, because sedimentologists
take sediment samples, and not populations.

Cheers,

Dr Matt Fischer
School of Geosciences
University of Sydney

Colin.Pain at ga.gov.au wrote:

> There would seem to be a more fundamental question, admirably posed by
> Robert Ehrlich in 1983 in Volume 53(1) of J. Sed. Pet. I reproduce it
> below.
> 
> ------------------------
> SIZE ANALYSIS WEARS NO CLOTHES, or HAVE MOMENTS COME AND GONE?
> 
> For a long time we have assumed that size frequency distributions
> contain a veritable treasure-trove of information. Mean (or median) size
> and measures of sorting have indeed become practical tools. Steadily
> over half a century papers have demonstrated that still finer nuances of
> distributions can be evaluated and so, of course, are of value. The
> present technological revolution in instrumentation and computers
> carries a threat that shortly we will be inundated by even more papers
> illustrating even more complex approaches which the authors will regard
> as Promising. Promises Promises. . . . Hope springs eternal. . . . "The
> check is in the mail." A troubling thought inkles its way to the fore.
> Why don't we routinely use this tool of great worth to help solve the
> sorts of problems that our field addresses? Have the research objectives
> that stimulated this approach slowly and unobtrusively evaporated
> leaving complex size analysis a cure that has lost its disease? If such
> objectives are still viable, why has no one over the span of more than
> 50 years stumbled on the rights "combination" to achieve them? One
> certainly can't attribute the lack of success to the intellectual level
> of practitioners inasmuch as many of our most honored colleagues have
> taken a hack at the problem at one time or another. The reason lies
> elsewhere. The time has come to re-evaluate the effort, try to diagnose
> the roots of this unsavory situation. Perhaps the time has come to stop
> plaguing generations of students with complex techniques that are never
> seriously used. It appears to me that their time would be more
> profitably engaged in tatting.
> ---------------------
> 
> 
> _______________________________________________________________ 
> Colin Pain, PhD
> Cooperative Research Centre for Landscape Environments & 
>     Mineral Exploration (CRC LEME)
> c/- Geoscience Australia   {Room 2.003}
> ABN 80 091 799 039 (GA)   ---   ABN  26 980 905 034 (CRC LEME)
> Postal address          Courier/visit address
> GPO Box 378                Cnr Jerrabomberra Ave & Hindmarsh Dr
> Canberra, ACT 2601     Symonston, ACT 2609
> Australia                       Australia
> Contacts
> Tel:      (02) 6249 9469 - International: 61 2 6249 9469
> Fax:     (02) 6249 9930 - International: 61 2 6249 9930
> Mob:     0409 225 008 - International: 61 0409 225 008 
> Email:    Colin.Pain at ga.gov.au
> Web:     LEME: [http://crcleme.org.au]   GA: [http://www.ga.gov.au]
> _______________________________________________________________ 
> 
> 
> 
> 
> _______________________________________________
> Aqualist mailing list
> Aqualist at anu.edu.au
> http://mailman.anu.edu.au/mailman/listinfo/aqualist
> 

-------------------------------------------------
This mail sent through IMP: www-mail.usyd.edu.au