Uncertainties and Reporting of Measurement Results: a Note of Caution for Authors

The availability of automated analytical testing systems and computerised data evaluation has dramatically increased the abundance of measurement data. Nevertheless, results of physico-chemical measurements invariably contain an element of uncertainty, which results from random variation (noise); on top of that systematic effects (bias) might also influence the analytical results. Numerical data for the expression of measurement results should take into account the degree of uncertainty by rounding them to just give the number of decimal places that are really significant (significant figures). Figures (digits) not carrying information should be removed; listing of tediously large numbers of digits makes the use of such data sets inconvenient and prone to transcription errors. The Editorial Board of EJLST noted that many of the manuscripts submitted report results in a way which is based on arbitrary reasons of habit rather than statistical consideration. As a general rule textbooks dealing with applied measurement science recommend to quote only digits known with certainty plus the first uncertain digit. For doing that rounding rules are necessary which let one allow to drop insignificant digits produced either by measurement instruments or derived by computations. At this point it is noteworthy to remember that rounding changes the original value and therefore introduces a new error. This source of error has to be controlled in such a way that rounding does not significantly enlarge the originally reported uncertainty. DIN 1319-3 [1] states: the value for the uncertainty is to be rounded off to two (or, where necessary, three) significant figures. Please see the whole abstract in the attachment.

ULBERTH Franz; 

2007-08-22

WILEY-V C H VERLAG GMBH

JRC36691

https://publications.jrc.ec.europa.eu/repository/handle/JRC36691,