I think it is a bit risky to state that Excel gives wrong results. We do not
know which version of Excel is being used by the original post author. I
think we should be presented with the mathematical formula, the formula in
the Excel cell and the exact numerical values in which the difference has
been observed. The difference may be due to rounding errors, the order in
which the calculations are made etc. Floating point operations are
approximate and TI-30 is very old (at least 25 years) so it may be due to
the algorithms used there.
If the way calculations in both cases are done correctly and in identical
way it is wotrth considering whether the differences are significant in
comparison to the statistical uncertainty. It may turn out that after
taking the uncertainty into account (its two significant digits), the
difference is insignificant.
--
Dr. Grzegorz Adamiec
Silesian University of Technology, Institute of Physics Department of
Radioisotopes, GADAM Centre of Excellence Krzywoustego 2, 44-100 Gliwice,
POLAND tel. +48 32 237 2696, fax. +48 32 237 2254 http://www.carbon14.pl/
------------------------------
Hello Tom,
The statisticians in my unit routinely warn us about Excel calculations, and
this goes beyond statistics even to graphics. They cite several articles
showing that Excel has problems, among which are:
(1)
<http://www.stat.uni-muenchen.de/%7Eknuesel/elv/excelxp.pdf>http://www.stat.
uni-muenchen.de/~knuesel/elv/excelxp.pdf
(2) D. McCullough B.; Wilson B. On the accuracy of statistical procedures
in Microsoft Excel 2000 and Excel XP.
<http://www.ingentaconnect.com/content/els/01679473;jsessionid=3nn4j0onoc271
.victoria>Computational
Statistics and Data Analysis, Volume 40, Number 4, 28 October 2002, pp.
713-721(9).
(3)
<http://gcrc.ucsd.edu/biostatistics/Excel.pdf>http://gcrc.ucsd.edu/biostatis
tics/Excel.pdf
This being said, I use Excel all the time for calculations and for graphs,
and suppose I will eventually be sorry.
Tom Brenna
At 10:56 AM 3/3/2007, you wrote:
>Hello All,
>We've been asked recently to provide results for a project in atom
>percent instead of the usual delta per mil notation. It took a little
>while to find the equations for this, so I decided to add one more
>place online where this could be done. What I found in the process is
>that if I convert delta per mil to atom percent using Excel, I get one
>result. If I use my old TI-30, I get a slightly different result. If I
>use php
>(http://localhost/uasil/black/isotopes/delta_to_atom_percent.php), I
>get the same result as with the TI-30. The results of the calculator
>and php are different from what I get with Excel starting with the 3rd
>to 5th decimal place, depending on the standard and isotope used. This
>degree of precision is normally adequate, but I'd expect better;
>particularly if I'm making a number of calculations. I'm assuming all
>three tools have adequate floating point precision to produce similar
results for more than just a couple of decimal places.
>Why would there be differences after the 2nd decimal place with only a
>few calculations (in the case of delta per mil to atom percent)? I'm a
>little curious (concerned) as we use Excel daily for much more
>complicated calculations.
>Thanks,
>-Tom
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>University of Arkansas Stable Isotope Laboratory
>http://www.uark.edu/ua/isotope/
--=====================_1347542468==.ALT
Content-Type: text/html; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<html>
<body>
Hello Tom,<br><br>
The statisticians in my unit routinely warn us about Excel calculations, and
this goes beyond statistics even to graphics. They cite several
articles showing that Excel has problems, among which are:<br><br> (1)<br>
<a href=3D"http://www.stat.uni-muenchen.de/%7Eknuesel/elv/excelxp.pdf">
http://www.stat.uni-muenchen.de/~knuesel/elv/excelxp.pdf</a><br><br>
(2) <font size=3D2>D. McCullough B.; Wilson B. On the accuracy
of statistical procedures in Microsoft Excel 2000 and Excel XP.
<a
href=3D"http://www.ingentaconnect.com/content/els/01679473;jsessionid=3D3=
nn4j0onoc271.victoria">
Computational Statistics and Data Analysis</a>, Volume 40, Number 4, 28
October 2002, pp. 713-721(9). <br><br> (3)<br> </font><a
href=3D"http://gcrc.ucsd.edu/biostatistics/Excel.pdf">
http://gcrc.ucsd.edu/biostatistics/Excel.pdf </a><br><br> This being said, I
use Excel all the time for calculations and for graphs, and suppose I will
eventually be sorry.<br><br> Tom Brenna<br><br> <br> At 10:56 AM 3/3/2007,
you wrote:<br> <blockquote type=3Dcite class=3Dcite cite=3D"">Hello All,<br>
We've been asked recently to provide results for a project in atom percent
instead of the usual delta per mil notation. It took a little while to find
the equations for this, so I decided to add one more place online where this
could be done. What I found in the process is that if I convert delta per
mil to atom percent using Excel, I get one result. If I use my old TI-30, I
get a slightly different result. If I use php (<a
href=3D"http://localhost/uasil/black/isotopes/delta_to_atom_percent.php"=
eudora=3D"autourl">
http://localhost/uasil/black/isotopes/delta_to_atom_percent.php</a>), I get
the same result as with the TI-30. The results of the calculator and php are
different from what I get with Excel starting with the 3rd to 5th decimal
place, depending on the standard and isotope used. This degree of precision
is normally adequate, but I'd expect better; particularly if I'm making a
number of calculations. I'm assuming all three tools have adequate floating
point precision to produce similar results for more than just a couple of
decimal places. <br> Why would there be differences after the 2nd decimal
place with only a few calculations (in the case of delta per mil to atom
percent)? I'm a little curious (concerned) as we use Excel daily for much
more complicated calculations.<br> Thanks,<br> -Tom<br><br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
University of Arkansas Stable Isotope Laboratory<br> <a
href=3D"http://www.uark.edu/ua/isotope/" eudora=3D"autourl">
http://www.uark.edu/ua/isotope/</a> </blockquote></body> </html>
--=====================_1347542468==.ALT--
------------------------------
Date: Tue, 6 Mar 2007 12:54:11 +1200
From: Peter Franz <[log in to unmask]>
Subject: Re: Microsoft Excel precision
I too once encountered a precision problem - it's been a while, but I think
it was either in the exp() or log() function. I pretty much nailed it down
to that - when I put my snippet of code into a Pascal file it worked without
fault, but in Excel/Visual Basic it just produced crap.
Peter
On Tuesday 06 March 2007 10:26, Tom Brenna wrote:
> Hello Tom,
>
> The statisticians in my unit routinely warn us about Excel
> calculations, and this goes beyond statistics even to graphics. They
> cite several articles showing that Excel has problems, among which are:
>
> (1)
>
><http://www.stat.uni-muenchen.de/%7Eknuesel/elv/excelxp.pdf>http://www.
>stat .uni-muenchen.de/~knuesel/elv/excelxp.pdf
>
> (2) D. McCullough B.; Wilson B. On the accuracy of statistical
>procedures in Microsoft Excel 2000 and Excel XP.
>
><http://www.ingentaconnect.com/content/els/01679473;jsessionid=3nn4j0on
>oc27 1.victoria>Computational Statistics and Data Analysis, Volume 40,
>Number 4,
> 28 October 2002, pp. 713-721(9).
>
> (3)
>
><http://gcrc.ucsd.edu/biostatistics/Excel.pdf>http://gcrc.ucsd.edu/bios
>tati
>stics/Excel.pdf
>
>
> This being said, I use Excel all the time for calculations and for
> graphs, and suppose I will eventually be sorry.
>
> Tom Brenna
>
> At 10:56 AM 3/3/2007, you wrote:
> >Hello All,
> >We've been asked recently to provide results for a project in atom
> >percent instead of the usual delta per mil notation. It took a little
> >while to find the equations for this, so I decided to add one more
> >place online where this could be done. What I found in the process is
> >that if I convert delta per mil to atom percent using Excel, I get
> >one result. If I use my old TI-30, I get a slightly different result.
> >If I use php
> >(http://localhost/uasil/black/isotopes/delta_to_atom_percent.php), I
> >get the same result as with the TI-30. The results of the calculator
> >and php are different from what I get with Excel starting with the
> >3rd to 5th decimal place, depending on the standard and isotope used.
> >This degree of precision is normally adequate, but I'd expect better;
> >particularly if I'm making a number of calculations. I'm assuming all
> >three tools have adequate floating point precision to produce similar
results for more than just a couple of decimal places.
> >Why would there be differences after the 2nd decimal place with only
> >a few calculations (in the case of delta per mil to atom percent)?
> >I'm a little curious (concerned) as we use Excel daily for much more
> >complicated calculations.
> >Thanks,
> >-Tom
> >
> >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >University of Arkansas Stable Isotope Laboratory
> >http://www.uark.edu/ua/isotope/
--
============================================================
Dr Peter Franz
National Institute for Water and Atmospheric Research (NIWA)
301 Evans Bay Pde
Private Bag 14901
Kilbirnie
Wellington
New Zealand
Ph : ++64-(0)4-386-0506
Fax: ++64-(0)4-386-2153
Mobile: ++64-(0)211-109304
Email: [log in to unmask]
Website: www.niwa.co.nz
============================================================
------------------------------
Date: Mon, 5 Mar 2007 16:48:28 -0800
From: Sam Ghidan <[log in to unmask]>
Subject: Working with Zn Natural fractionation
--0-1656316628-1173142108=:10981
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Dear Colleagues.
if any one works to detect variations in the isotopic composition of Zn
using the Double spike technique, then please just let us have more
discussion about it.
Variation in the isotopic composition of Zn is too small(<1 per mil) and
it is too difficult to detect due to interferences accompanied with the
measurements. the main interference is the interference in the 67 isotope of
Zn.
Thanks
Sam Ghidan
|
