Abstract: In this paper,a new algorithm is put forward as an alternative to the existing crespondence analysis algorithm in multivariate statistical inference,which differs from the existing algorithm mainly in the following two aspects: instead of ,and (2) performing necessary data pre-processing through the transformation formula and the constraint x'_i.=1 (or other constant).Moreover,a practical experimental example is given to make a brief comparison between the here-stated algorithm and the existing one as well as the indi-vidual R-mode and Q-mode factor analyses.Both theory and example prove that the algorithm stated in this paper is more logical in theory and applicable in practice than the existing one and overcomes the limitation of various kinds of geo-data on the existing one.