Abstract
Monte Carlo simulation studies are used to compare the classical (parametric) procedures with the rank transform procedures in the two by two factorial experiment. The rank transform procedures involve using the classical procedures on the ranks of the data. In this paper, it is demonstrated that the magnitudes of the main effects and interaction can have a pronounced effect on the rank transform procedures. When no interaction is present the two procedures (parametric and rank transform) are in close agreement when the error terms have a normal distribution. Results of Monte Carlo studies with error terms having various distributions are also presented and interpreted. It is shown that the rank transform procedure is superior to the parametric procedure when the distribution of the error terms is heavy-tailed like the double exponential.