Building robust regressions involves reducing the influence of one or more anomalous values on the calculation of the parameters of a mathematical model. Anomalous values are adjusted in one of three ways: discarding an outlier or an anomalous value, reducing the weight of this value, or refining the value through additional statistical analysis. Approximation of empirical data when building the mathematical model using the ordinary least squares method is not always optimal, so today there is a need to consider and analyze alternative approaches. This paper is devoted to the issues of building robust linear regression with outlier correction for samples with small sizes. The described technique was demonstrated using a specific numerical example. The final regression model was obtained by finding the geometric mean value of the tangent of the angle of inclination for three considered methods: the ordinary least squares method, the least absolute deviations, and the least range cumulative curve deviations.