Abstract
The performances of smoking machines of different types cannot be compared directly if different numbers of cigarettes are smoked into one trap, because the respective variabilities are depending on the number of cigarettes smoked into one trap. For the comparison of means, the t-test can therefore not be used directly. It is demonstrated that a generalized t-test can notwithstanding be applied to this problem in most cases, in particular when the standard deviations per cigarette are of the same order of magnitude