ABSTRACT
The Chebyshev method leads to wide confidence limits, whereas the Gaussian fit may lead to severe systematic error. Identification also for other distributions can be useful, because often the circuit performances do not follow a normal distribution. Leakage current follows often an exponential law, so assuming a lognormal behavior is much more meaningful than assuming normal data. The sum of normal variates is again normal, but actually the sum of two uniform variables is giving a triangular distribution, and the sum of two lognormal variables is not lognormal. Non-normal methods are needed because circuits really create such nonnormal distributions more or less all the time.
