This paper describes how sequential statistical testing, which usually is applied in an automated response detection algorithm, is time efficient but unfortunately also increases the probability of a false rejection of the null-hypothesis. Therefore, in such test situations the test criterion is normally modified by means of the Bonferroni correction. However, when dealing with dependent or partly dependent data the Bonferroni correction will lead to an over-correction and will therefore not be optimal. A new method to find the optimal test criterion is devised and tested by means of Monte Carlo simulations using real background noise data acquired from clinical ASSR-recordings.
