Trans. IEE of Japan, Vol.111-A, No.11,
pp.993-998 (1991.11)
Abstract
It has been indicated that an estimate S of the standard deviation
S obtained by the up-and-down method has some bias and the method
for determining the confidence interval of S through the Fisher's
amount of information is inaccurate. This paper shows that the
bias of S can be corrected by using the bias corrected bootstrap
method and that an accurate confidence interval of S can be obtained.
For easy understanding, a detail explanation of the method for
determining the confidence interval is shown in this paper by
a concrete example of the up-and-down method. This is an introduction
for the practitioners. This method was proved correct by the Monte
Carlo simulation. The up-and-down method is still popular as a
testing method for determining the impulse-breakdown voltage.
This method is particularly effective when the estimate of the
standard deviation S of the breakdown voltage is larger than the
up-and-down voltage distance d (i.e., when d/S is small).
The paper written in English is available
upon request.