Abstract
The present paper discusses some robust statistics for psychologists. In a Monte Carlo study the efficiency of some robust alternatives to the mean and the standard deviation are studied in samples of varying sizes and varying degrees of deviation from normality. The median and the interquartile range appear to be insufficiently efficient under the normal model, while Gini’s difference and a related estimator of the population mean seem to be not only robust under deviations from normality but also fairly efficient in normal samples. Furthermore, the usefulness of a specific robust statistic for estimating the population product-moment correlation is discussed. The article stresses the importance of robust statistics that are routinely applicable and quite efficient under conditions of normality. Some suggestions for statistical inferences based on these statistics are offered.