Sediment contaminant concentrations usually show an inverse correlation with grain size. This can cause difficulties in distinguishing real differences in contamination from artifacts caused by variations in sediment texture. To overcome this, regression analysis is frequently used to remove the dependency of concentrations on grain size. However, least squares regression lines can be affected markedly by the presence of a small number of unusual samples in the dataset. These outliers may represent samples which are more severely contaminated or which were derived from areas with different underlying geology. They can be removed semi-manually, but robust regression methods such as least absolute values provide a convenient and objective alternative. The methods are illustrated using an example dataset of metal contaminants in sediments from the Hunter Estuary, United Kingdom. Least squares regression on the complete dataset yields a rather poor grain size normalization for several elements. By contrast, least absolute values regression produces results very similar to those obtained by least squares regression after careful manual removal of outliers, but it avoids the need for subjective judgments of which data points to omit from the analysis. The intercepts of several of the fitted regression lines were non-zero, indicating that regression-based normalization is preferable to methods based on ratios.