-----Original Message-----If one were to round off a result, one would also add to the likelihood of error in the result.
From: Les Culank [mailto:[log in to unmask]]
Sent: 19 November 2003 14:37
To: [log in to unmask]
Subject: Re: Significant figures, rounding off, and error
For example if all results from 9.5 to 10.4 were rounded off to 10, the result of 10 would gain in error, with a cv of approx. 3% from the rounding off process.
If the analytical error from all other sources were about 2% or so, the overall error of that result of 10 would now be approx. 4%.
Best wishes,
Les
From: Roger Ekins <[log in to unmask]>
Reply-To: Roger Ekins <[log in to unmask]>
Date: Wed, 19 Nov 2003 13:04:41 +0000
To: [log in to unmask]
Subject: Re: Significant figures
Dear oh dear!
This discussion caught my eye whilst waiting for an overdue visitor.
From a formal point of view, Jonathan's way of looking a things is not only simplistic - it is incorrect. A result of 127 with an s.d. of 0.2 should be reported as 127.0 ± 0.2. In other words the first decimal place is significant. Likewise a result of 126.5 with a 2% error should be reported as 126.5 ± 0.2 - or 126.5 ± 0.3 if one wishes to be on the safe side. After all, how accurate is the estimate of the sd?
However as I understand John Beilby's question - if I understand it at all - he and his colleagues seem to be debating whether reporting a value of 127.5 is necessary when all the clinician may wants to know is: is the value >100 and <150?.
Why clinical biochemists should be debating this issue puzzles me. Their job is to report the result of an analyis, and in my view - though I'm not a clinical biochemist - they should report results of analyses including all significant figures - ideally with an error attached as above. They should not presume knowledge regarding who will use the result and how it will or might be used.
On the other hand if it is certain that a result with an sd of 1 meets all known clinical requirements, then the clinical chemist may be wasting time and money in using instruments or methods producing results characterized by errors of 0.1, 0.01 or whatever. In other words the methodology used should such as to meet clinical needs. There's no point in using a micrometer to measure the length of a fish. But this is not the same question as that raised by John Beilby.
Also it is irrelevant to this issue whether results are reported as mmol/L or µmol/L. More convenient of course to use units that yield mostly whole numbers.
Roger Ekins
Molecular Endocrinology
University College London
Hi
The way I (simplistically) look at this issue is to consider the implied uncertainty of the measurement.
If you have a creatine result of 127 micromol/L, this implies that you can state with confidence that the result lies between 126.5 and 127.5 micromol/L. If 0.5 micromol/L therefore represents 2 SDs of the repeat measurement imprecision, 1 SD is 0.25 micromol/L, so the CV is 0.25/127 x 100 = about 0.2% (if this is the only source of uncertainty)!
If your uncertainty is nearer 2%, then rounding up to the nearest 10 (ie 130) is probably more appropriate. If your uncertainty is 20% then 'about 100' is probably all you can say.
By making this response I probably open myself up to a deluge of criticism from statistical quality control experts, but I hope this is helpful!
Cheers
Jonathan Middle
-----Original Message-----
From: "Beilby, John" <[log in to unmask]>
To: [log in to unmask]
Date: Wed, 19 Nov 2003 18:19:05 +0800
Subject: Significant figures
Dear All
We have recently had considerable discussion concerning the number of
significant figures that should be used in reporting biochemistry results.
Many laboratories in Western Australia report serum creatinine results in
umol/L, while other parts of Australia report in mmol/L.
It has been suggested that if we are going to use umol/L units we should
only report to the nearest 10 umol/L. I am interested in how other labs
deal with this problem as we cannot find much in the literature except for a
letter by Badrick and Hickman in Ann Clin Biochem 2002;39:162.
If we adopt the argument in this letter how many significant figures should
be used to report glucose and urea results, for example.
John
Dr John Beilby
Senior Biochemist
Clinical Biochemistry
PathCentre
Locked Bag 2009
Nedlands
Western Australia 6009
Tel 61 8 9346 2368 Fax: 61 8 9346 3882
[log in to unmask]
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical
community working in clinical biochemistry.
Please note, archived messages are public and can be viewed
via the internet. Views expressed are those of the individual and
they are responsible for all message content.
ACB Web Site
http://www.acb.org.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
UK NEQAS Birmingham
tel 0121 414 7300, fax 0121 414 1179
This message is intended only for the above named recipient(s). Message content may be confidential and privileged. If it comes to you in error please inform the sender and delete it from your system.
The opinions expressed are mine alone and do not necessarily represent those of UK NEQAS Birmingham, the UK NEQAS Organisation, the University Hospital Birmingham NHS Trust or University of Birmingham.
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical
community working in clinical biochemistry.
Please note, archived messages are public and can be viewed
via the internet. Views expressed are those of the individual and
they are responsible for all message content.
ACB Web Site
http://www.acb.org.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
--
------ACB discussion List Information-------- This is an open discussion list for the academic and clinical community working in clinical biochemistry. Please note, archived messages are public and can be viewed via the internet. Views expressed are those of the individual and they are responsible for all message content.
ACB Web Site http://www.acb.org.uk List Archives http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html List Instructions (How to leave etc.) http://www.jiscmail.ac.uk/
------ACB discussion List Information-------- This is an open discussion list for the academic and clinical community working in clinical biochemistry. Please note, archived messages are public and can be viewed via the internet. Views expressed are those of the individual and they are responsible for all message content.ACB Web Site http://www.acb.org.uk List Archives http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html List Instructions (How to leave etc.) http://www.jiscmail.ac.uk/