Het quantitatieve verband tussen dosering en letaliteit bij proefdieren

Summary (On the quantal response of animals to drugs). The methods of Bliss (6) and Fisher to deal with dosage mortality data were explained till the computation of the final regression line, which will be treated in a following article. These methods are based on the tacit assumption that the dosage mortality data do follow a constant probality distribution. This is not always true. In the cases of t.b.‐infected Cavias (7) and the convulsive effect of insuline to mice (4) the experiments are irreproducible and so a factor F must exist, which varies in an unknown and irregular way between the series. It is still possible then, to draw trustable inference relative to, say, two different preparations A and B of the drug, by mixing up the two series completely at random, safe for the difference between A and B. In this way F is degraded to an ordinary random variable and the analysis may be executed, as far as the difference goes, as if F were nonexistent. The logarithmic distribution of the susceptibility to drugs may be taken as een instance of a wider rule, which embraces also the facts collected in the law of Weber—Fechner and ought to be called the Rule of Galton: In biology the logarithmic progressions are predominant to the arithmitic ones. That the logarithmic progressions are peculiarly obvious in the cases of susceptibility to drugs or to sensual stimuli, is due to the fact that here the range of variability is relatively large. Apparently the change of scales takes place at the very moment in which the physico‐physiologic frontier is trespassed. In the eye the current of action in the retina is closely proportionate to the logarithm of the physical magnitude of the light. According to Bliss (6) the adsorption in the tissues should be proportionate to the logarithm of the dose, rather than to the dose itself. It must however be said that the coincidences, shown by Bliss of log dosis‐probit‐curves, departing somewhat from normality, and suitably selected fragments of Langmuir‐sigmoids, are little convincing evidence for his views, because of the large flexibility of the sigmoids (3 parameters). The keen interest taken by Bliss in departures from normality of dosage mortality data is comparable to the interest taken in departure from normality at large in the childhood of biostatistics. It was seldom forgotten then, to compute the skewness and the kurtosis, or better still, to fit the best‐suiting Pearsonian curve, of any distribution. Much more seldom however, anything was done with the results. Nowadays the value of departures from normality is considerably depreciated. Significant departures from any normal or Pearsonian curve may always be shown, by taking enough observations, because not a single natural distribution can be exactly identical with one of those curves.