Paleonet: histogram showing bimodal distribution: A hint regarding the Excel-glitch Prof. Campbell mentioned

Peter Smolka PeterPaulSmolka at T-Online.de
Thu Aug 23 10:15:18 GMT 2012


The reply added below

-----Ursprungliche Nachricht-----
Von: paleonet-bounces at nhm.ac.uk [mailto:paleonet-bounces at nhm.ac.uk]Im
Auftrag von David Campbell
Gesendet: Mittwoch, 22. August 2012 17:05
An: PaleoNet
Betreff: Re: Paleonet: histogram showing bimodal distribution


Be aware that Excel does not recognize the existence of bimodal
distributions in its statistical tools.  If there are multiple modes,
it picks one without any evident notification that there are others.
In general, it is much safer to enter the statistical formulas you
want to use than to trust Excel's built-in statistic tools (also true
of any matrix algebra).  Of course, for serious statistics, you're
probably using a statistics program anyway, but this can be an issue
for, e.g., a basic-level lab class.

--
Dr. David Campbell
Visiting Professor
Department of Natural Sciences
Gardner-Webb University
Boiling Springs NC 28017


Reply:

The safest way:

Davis,  : Statistics and Data Analysis in Geology, the OLD edition 1973 or
so.

They have the programs in full transparency in Fortran. You just can type
them (and edit them in some details)
and have full control over what you are doing.

A safe free (for non-commercial use) Fortran compiler: The Salford compiler.

The Salford-Compiler finds non-initialized variables at EXECUTION-time by
the /checkmate -option (most others set variables to zero in such
case and continue just to produce wrong results).

Intel, after an explanation on that by me about in the year 2000, added it
stepwise into their Fortran Suite (the /check:all option).
I wrote: NO program that is not tested with /checkmate (or an equivalent)
shall be regarded as correct.

So it works pretty fine, also, as I understood, regarding other languages
Intel offers (which I regard as setback, e.g. C and such).

Of course one can hope that technical software, such as by Airbus, is also
tested with the /check:all option (Salford: /checkmate )
and equivalents.

For those who know it: /check:all and /checkmate are overall the equivalents
of UV-0 of WATFIV and UV-05 of WATFOR.

So: To have the control Prof. Campbell mentioned I recommend the "old"
Davis. The "new" Davis does not have the transparency, e.g.
the source ready for typing.

Best regards

Peter

P.S.: Above is a _private_ reply to the question on paleonet.

Email:
PeterPaulSmolka at T-Online.de (Private)
The other email exists of course.

And for those that are interested in the dangers of C:

There is an article by E. Sturm (physicists, computer-center of University
of Muenster) who presented it at a SHARE conference of IBM, somehow with
the title "Chaos vs. Reliability" (in english). One of the various
(breathtaking) glitches of C: Under SOME circumstances operaters can change
their meaning
unintended(!) - with F66 / F77 this never happens.

In-line with the issue of Prof. Campbell (correct results) I included this
hint.
_______________________________________________
Paleonet mailing list
Paleonet at nhm.ac.uk
http://mailman.nhm.ac.uk/mailman/listinfo/paleonet





More information about the Paleonet mailing list