Response to “The Statistics of Skin Impedance”

Dear Editor:

We extend our thanks to Quintanilha et al. for raising a pertinent issue. ¹ Indeed, it has been observed that skin impedance generally follows a log-normal distribution. Hence, using a log transform would probably have helped our data to resemble a normal distribution. However, the issue of applying transformations to skin impedance data is still controversial. Bouscein, in his comprehensive text on electrodermal activity (EDA), notes that transformations used in EDA research are seldom based on physiologic or systems-theoretical considerations. ² Rather, he suggests, statistical considerations lead to the transformations of electrodermal raw data. He recommends that transformational procedures be judged by the analysis of the derived measure, whether or not the procedure leads to an improvement of the validity of the electrodermal measure. The non-normal nature of our data necessitated the use of nonparametric measures. ³ Thus, we used a conservative approach so as to not inflate type I errors in the case of such distributions.