Environmental data often presents censored, lost and/or outlier values. In addition, samples should be considered dependent for having spatial and temporal components. Another fact is that, frequently, these data won't follow a Normal or Log-normal distribution. Because of these and other characteristics, conventional statistical techniques should not be used. This article presents a case study of the Das Velhas river, Minas Gerais, using robust statistical methods after appropriate treatment of the data. The analysis of the main components found the variables that contribute the most for the degradation of water quality in the river, and the spatial visualization of the scores showed where this contamination is most evident.
robust statistics; environmental data; principal components analysis; R software; Das Velhas river