In NSVOK, the Robust Algorithm is applied for statistical comparison:
(reference ISO 13528:2005(E) - Annex C (Robust analysis)
Determination of starting median
x* = median of xi (i = 1, 2, ..., p)
An initial standard deviation is determined
s* = 1.483 median of |xi − x *| (i = 1, 2, ..., p). ϭ=s*.1.5
Remove grossly erroneous results
For each measured value
(xi): > max xi=( x* + ϭ) are equated to max
xi < min xi=( x* - ϭ) are equated to min xi
With these values of xi are calculated
x* = Σ xi/n
s* = 1.134 √ [ 1/(n-1) * Σ (xi - x*) ]
The new s* thus obtained is multiplied by 1.5 to obtain the new ϭ
ϭ=s*.1.5
The calculation is repeated with the new values of max xi and min xi , until the first three digits of x* and s* become the same as those of the previous calculation.
The resulting values are:
x* = target value(X)
s* = SD
Each participant's z-score, CV% and D% are calculated based on their measured values.
z-score(SDI) = [X-x]/SD
CV% = SD*100/X
D% = 100*[(x-X)]X