PAPTAC - Z.2P
ACCURACY AND PRECISION OF STANDARD METHODS
|Publication Date:||1 November 1981|
In general, the random variations affecting test methods are normally distributed (i.e., give a Gaussian frequency distribution curve) and therefore the computations are based on this distribution. In case of non-normal distributions consult reference 2 or 3. Random variations are those which arise from small and unavoidable errors in such operations as weighing, reading instruments, etc., and which over a number of replicate measurements, tend to average out to zero. They are not to be confused with systematic error or bias - as typified by improper calibration of apparatus - which does not reduce precision but decreases accuracy. Bias is the most dangerous form of error and special care must be taken to avoid it when devising sampling schemes.