Simple method for outlier detection in fitting experimental data under interval error

A method for outlier detection in building linear regression for data with interval-bounded error is described. The core idea of the method is as follows. An outlier may be treated as an observation with the underestimated error i.e., an actual error interval is wider than the declared one. In order to correct an outlier it is necessary to find out the error lower bound which makes the observation consistent with the others. Comparison of the value of the lower bound of the possible actual error to the value of the declared observation error allows to make some inferences concerning the degree of inconsistency of an outlier to the entire dataset. Lower bounds of actual errors are determined as a solution of a linear programming problem. The proposed technique is applied to solve the problem of the geometric correction of satellite images.