Dr Rory O'Conor
Consultant Clinical Epidemiologist
Clinical Audit
Pinderfields & Pontefract Hospitals NHS Trust
Wakefield WF1 4DG, UK
Tel: +44 1924 212850
Fax: +44 1924 814864
Email: [log in to unmask]
X400:"c=gb;a=nhs;p=nhs n and y hn;
o=nhs pinderfields and pontefract hospitals tr;
ou=gwise;s=oconor;g=rory"
http://nww.panp-tr.northy.nhs.uk/ (extranet)
http://www.demon.co.uk/pinderfields/ (internet)
http://www.btinternet.com/~rory.oconor/ (personal)
>>> DDA.RFC-822=evidence-based-health-request(a)mailbase.ac.uk 27/05/99 17:32 >>>
Even away from the singularity, the ARR and NNT, though they contain
equivalent information, look rather different mathematically. I
wouldn't advocate bandying about small decimals like 0.00235,
but 2.35 per 1000 seems assimilable. A fair degree of rounding is
sensible, when professionals talk about study results informally
among themselves, but the definitive documentation should give things
to "about two-and-a-half" significant figures, whichever scale is
used. (I can't give a reference for this quote, I'm afraid.)
..........................................
Robert G. Newcombe, PhD, CStat, Hon MFPHM
Senior Lecturer in Medical Statistics
The number of significant digits is an indication of the precision of a measurement. It may be entirely appropriate to record pi, force of gravity, speed of light to 5, 10 or even 50 significant digits.
With statistics such as 2.35 per 1000, the degree of precision may be 1-10 per 1000, 2 per 1000, 2.35 per 1000 depending on the power of the dataset used and number of events studied. Thus rounding is used to demonstrate the power of the data.
Obviously when a result is going to be reused in further calculations then it is better to use the original unrounded figures and these may contain 1-2 more significant figures than the precision warrants.
It is easy with a spreadsheet, where you can display everything to an appropriate level of precision, whilst preserving the underlying original result.
Rory
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|