Statistics for analytical chemist

[pi3832]

My brain ain't what it used to be.

I could use a good referesher on the more archane bits of analytical statistics, specifically error calculations and propogation of error.

Can anyone recommend a good stats book that's geared for analytical chemistry, or, better yet, GC? And not as gratuitously obfuscated as so many statistics books seem to be?

(Authors of statistics references seem to be obsessed with the most complicated and exceptional cases. Yeah, I know statistics is rigorous and can be applied to most anything, but must you prove it on every page?)

[MikeD]

Michael

There may be several books with titles such as "Statistics for Analytical Chemists" or similar but this is one of those occasions when I would recommend guides by committees.

The master document is the ISO Guide to Uncertainty of Measurement (GUM). In the US it was adopted as ANSI/NCSL Z540-2-1997. In Europe it's EN 13005:1998. It cost money, it's a tough read and probably no "guide" was less helpful to the beginner, so I will recommend three online guides to the guide by NIST, Eurachem and Nordtest.

NIST anticipated the GUM and published NIST Technical Note 1297,1994 Edition
Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results
http://physics.nist.gov/Pubs/guidelines/contents.html

The Eurachem guide is at http://www.eurachem.ul.pt/guides/QUAM2000-1.pdf

The Nordtest report TR537 is a very practical guide written for environmental test laboratories and is at
http://www.nordicinnovation.net/nordtes ... tec537.pdf

The Eurolab reading list is at http://www.eurolab.org/docs/top%20reading%20list.htm

[mbulentcakar]

hi,
shaun burke wrote some very good articles about statistics for lcgc europe. they may be of great help.


http://www.lcgceurope.com/lcgceurope/da ... rticle.pdf
http://www.lcgceurope.com/lcgceurope/da ... rticle.pdf
http://www.lcgceurope.com/lcgceurope/da ... rticle.pdf
http://www.lcgceurope.com/lcgceurope/da ... rticle.pdf

[WK]

These articles are very good. Well thought out examples.
(I thought I saw an obituary in Chemistry World for Shaun Burke and he was quite young. He worked at LGC and RHM. Can anyone confirm? Most sad if so.)

Another good book is Statistics and Chemometrics for Analytical Chemistry by Miller and Miller. I have the 4th edition.

WK

[pi3832]

MikeD and mbulentcakar:

Thanks for the excellent links. I've downloaded the PDFs and am starting to go through them. Interesting how the European stuff seems so much more clear than the U.S. stuff.

BTW, I did find what appears to be a useful NIST website:
e-Handbook of Statistical Methods

Unfortunately, I've been finding that site a little hard to comprehend. At times it seems that some people make statistics more confusing than it needs to be.

(Anecdote:
A lab where I worked many years ago had this odd, intricate calculation that I was required to do on certain samples. For a while I just sighed and did it.

But then I started to dig into the procedure and ask where the calculation came from, and discovered that it was mostly bogus. Talking to others I eventually figured out that the guy who had written the procedure didn't know anything about statistics, so he had just taken some basic equations and algebraically manipulated them until they were unrecognizable and suitably complicated to look impressive.

I worked them backwards to their original form, then simplified them based on the limited application for which we were using them, and came up with a one-step calculation you could do on a calculator, instead of the previous one that had required a computer spreadsheet.

I've been dubious of complex statistical calculations ever since.)

[tx]

Interesting. I would like to know the statistic backgroud of setting up a limit for the agreement between two duplicate testing results. For example, when to accept or reject a testing result which is the average of two sample preparations. The criteria should be calculated from validation data, instrument repeatbility, method accuracy... But how?