Adjustment of Conventional Regression Methods by Using of Goal Programming Models and it's Application in Human Sciences

Inspite of the wide spread use of regression models in most areas of social sciences, the necessary prerequisities of them less be noticed and often are uncontrolable in these areas. The aim of this article is introducing a substitute for the Least Square regression (LSM) models by using of mathematical bases of goal programming and optimization techniques, with less sensitivity against of non-normal distributions and especially more robustness for outlier data. In the following of introducing the mentioned model called Least Absolute Values (LAV) regression, two models tested on data of a rural development study of influential factors on rurals participation in developmental projects in Kermanshah city, by using of SPSS and LINDO softwares respectively. Results showed the more precise estimations in LAV approach for independent variables cofficients compared with the LSM. Also the residuals of the LAV model estimations are 0 at 10 cases of 12 studied villages, whereas no one of the LSM estimates are 0. So, with the better use of collected data, the LAV approach leads to the better and nore accurate estimations.