Molecular weight and log P remain the most frequently used physicochemical properties in models that predict skin permeability. However, several reports over the past two decades have suggested that predictions made by these models may not be sufficiently accurate. In this study, exploratory data analysis of the probabilistic dependencies between molecular weight, log P and log Kp was performed on a dataset constructed from the combination of several popular datasets. The results suggest that, in general, molecular weight and log P are poorly correlated to log Kp. However, after employing several exploratory data analysis techniques, regions within the dataset of statistically significant dependence were identified. As an example of the applicability of the information extracted from the exploratory data analyses, a multiple linear regression model was constructed, bounded by the ranges of dependence. This model gave reasonable approximations to log Kp values obtained from skin permeability studies of selected non-steroidal ant-inflammatory drugs (NSAIDs) administered from a buffer solution and a lipid-based drug delivery system. A method of testing whether a given drug falls within the regions of statistical dependence was also presented. Knowing the ranges within which molecular weight and log P are statistically related to log Kp can supplement existing methods of screening, risk analysis or early drug development decision making to add confidence to predictions made regarding skin permeability.