Combined regression models (Q2430222)

Three linear regression fitting criteria are considered: the least squates (LS), least absolute deviations (LSAD) and minimax (MM). The authors consider estimates which can be obtained by minimizing one of these criteria when some other (or both) are bounded by some fixed number. Linear programming algorithms are proposed for calculation of these estimates. Assessment of the observations influence on the estimates is discussed. An example of a biological real life data set analysis is presented.