Zeroes, Missings, and Outliers

The presence of irregular data, i.e., zeroes, missings, and outliers, has consequences for plotting, descriptive statistics, estimation of parameters, and statistical tests. Since all of these methods depend on every value of the dataset, we cannot ignore them in any task. Some ad hoc procedures are needed, with the same aims as the classical methods, but which can still be computed despite the existence of irregular data. It is necessary to augment the basic concepts (e.g., the concept of expected value) to give them a meaning when there are irregular values. The current state of the art of the treatment of irregularities in compositional data analysis is far from being a closed subject and can improve a lot in the near future. The package only provides a set of tools limited to detect, represent, and briefly analyze such irregular values, either missing values, zeroes, or outliers. This chapter provides nevertheless additional background material to enable the reader to carefully treat datasets with irregular data.