“…19 34.4 ± 0.6 100 mm × 10 mm 332 ± 14 1.03 ± 0.07 32.5 ± 1.3 determination of the strength of materials at extremely high probabilities of failure is almost impossible (unless huge numbers of samples would be tested), enhanced alternative methods to extrapolate the statistical data towards the upper tail have to be used. As the number of individual tests carried out to characterize material strength is usually rather small, it is often difficult to choose among similar distributions to describe statistically this data, such as, for example, normal [16], log normal, Weibull and gamma, when only considering goodness-of-fit tests [17]. However, extrapolations based on these distributions greatly differ at their upper tail.…”