Abstract
The parameters.of the space influence model estimated by minimum square error method were unstable caused by the multicollinearity. A practical method for the robust estimation of the space influence model is proposed in this paper. Introducing the orthogonal variables W by using the principal component analysis, the space influence model can be transformed to the multiple regression model with form ; Y = WB. The parameters of the space influence model can be obtained by returning the regression coefficients B. For reducing the effect of the multicollinearity, (1) we omit the elements of B corresponding nearly zero eigenvalue. (2) we omit the elements of B corresponding variables W which have negligible small correlation coefficients with the dependent variable Y.
