Abstract: To quantitatively investigate the stability and authenticity of the statistical correlation coefficient, we propose a pair-wise deletion correlation coefficient (PWDCC) testing method which is based on χ2 test, through pair-wise deleting elements of a data sample and constructing a correlation coefficient series. The analysis using both ideal data sample and real climate data confirms the validity and feasibility of the PWDCC. The results show that, this method can objectively and quantitatively determine whether there exist extreme values which lead to large variation of the sample correlation coefficient. Furthermore, the method is simple and has clear physical significance, and it also overcomes some weakness of the traditional methods (the scatter diagram, the slide correlation method, etc).