ABSTRACT
Hence, the burning cost is, with the same probability, within the interval
XJS^f^XJS. (6.4.3)
Assume i n the above example that the fo l lowing numerical parameter
values are applicable ((6.1.10), (6.1.11) and (6.1.11a))
n = 5000, r2 = 40, r 3 = 4000, aq = 0 ,1 , yq = 0.5, e = 0.025. (6.4.4)
The observed mean c la im size is X/n = 0.00160, wh ich can be used
as an estimate for the mean m. Fur thermore , the standard deviat ion
a x and the skewness yx can be calculated from (3.2.15), obta in ing
the values 1.07 and 0.37 respectively. Then, apply ing the inverse
W H - f o r m u l a of section 4.2.5(b), the fo l lowing numerical result is
obtained for (6.4.3)
6.09/7000 = 0 .87% 10.28/7000 = 1.47%o. (6.4.5)
U s i n g the no rma l approximat ion , i.e. tak ing
Xi,2 = Vx±yEGx, (6.4.6)
the limits wou ld be 0.84 and 1.44 per thousand. The difference from the
above estimate is due to the fact that the normal approximation ignores
the effect of skewness.
