||The goal of this thesis is to make a reader familiar with the essential findings and techniques of the robust regression methods. Their undoubted advantage comparing to classical estimates represented mainly by the method of ordinary least squares is their ability to cope with contamination of the data. Because of the process of the economic transformation, contamination or more likely data heterogeneity is inherent to the data emerging from transition economies. The regression model of the Czech export into the EU during 1993 – 1999 period is used to demonstrate that robust estimates namely the least trimmed squares and the least median of squares achieve better characteristics than estimates acquired by the method of ordinary least squares. Moreover, it turns out that data heterogeneity detected by robust techniques has its economic theory based logic.