Multicollinearity Problems and Their Treatment in Statistical Studies of Land use Change

In the statistical analysis of Land Use/Cover Change (LUCC), it′s very common that the number of independent variables is not quite large yet the multicollinearity among them is very serious. We introduced the All possible regression method to resolve this kind of problem on the basis of summarizing the collinearity diagnosis methods and analyzing the advantages and disadvantages of stepwise regression method, which is now commonly employed among internal scholars to eliminate collinearity. A research on the LUCC driving forces in Baoan District, Shenzhen, was taken as an example. Using the All possible regression method, we successfully avoided multicollinearity among independent variables and established the explaining and simulating model of LUCC in Baoan. The model results are generally fine. Contrast study shows that the All possible regression method is a more effective method to eliminate collinearity in the statistical analyses of LUCC mechanisms.