My own rule of thumb, and perhaps a slightly different way of looking at it, is
that the probability that two results are different is <70% at a 1.5sd
difference, <90% at 2 sd difference and 95% at 2.8sd.
Peter Auld
______________________________ Reply Separator _________________________________
Subject: Re: AW: Significant change
Author: [log in to unmask] at XINTERNET
Date: 11/30/00 11:29 AM
re:-
>In Tietz Textbook of Clinical Chemistry 3rd ed. p288 the formula for the
>95% confindence interval of a true change is:
>(x1-x2) +/- 2 * root[sqr(sd1)+sqr(sd2)]
>(x1-x2) +/- 2 * sd * root[2]
>
>i.e. +/- 2.7 times standard deviation!
I seem to recall from the old days of Imperial umits, and longhand square
root extractions, that 2 * root2 was equal to 2.818 (to 3 dp).
Dr Les Culank, Consultant Chemical Pathologist
01223 - 217153 / fax 216862
Dept of Clinical Biochemistry, Box 232,
Addenbrooke's Hospital, Cambridge CB2 2QR UK
|