Fig. 2.18. Approximation by normal distribution. 
i.) p% Gaussian range of percentage error 0(w) 
BOS aie A. 
5) a) (2.103) 
1/2. 
where v(w) = [var. {s(@)}] : standard deviation 
bw) = E{s(w)|— s(w): bias 
c(p) is the two-sided p% point of the standardized normal distribution, 
< se) Se (2.104) 
Is(w) — s(w)I 
100 
S(@) 
ii.) The mean square percentage error 7(w) 
[zs — s(w F] _ V?@)+P?@) 
= 2.105 
2) s*(w) ae 
n°(w) = 
By Chebycheff’s inequality, 
39 
