Dealing with Nonnormal Data: Parametric Analysis of Transformed Data vs Nonparametric Analysis

Abstract

Researchers have typically employed parametric analysis of raw data to test experimental data for statistical significance. When the data are not normally distributed, data transformation or nonparametric analysis are often recommended. The present study compares parametric analysis of raw data to parametric analysis of transformed data and to nonparametric analysis when the tests are carried out under population nonnormality. The results of a Monte Carlo simulation indicate that when distributions depart markedly from normality, nonparametric analysis and parametric analysis of transformed data show superior power to parametric analysis of raw data. Furthermore, under the conditions studied, parametric analysis of transformed data appears to be somewhat more powerful than nonparametric analysis.