Multi-dimensional AR Series Modeled by Least Square Criterion

Being one of the modern methods of data processing,time serials analysis has a outstanding position in system identification and analysis.Based upon mutuality functions of probability statistics,traditional time series analysis can be completed only by complicated iterative method with non-linear least squares algorithm,whose theories and applications take only one factor into account at present,which is one-dimensional time series.Hence,it can not cater for dealing with practical multi-factor problems.Based on the least squares criterion widely used in surveying data processing,time-field analysis method of multi-dimensional AR series is illustrated in detail in this paper,which includes parameter estimations by the least squares algorithm and rank confirmation by routine F-test.The feasibility of the multi-dimensional time series analysis method is tested in the application of deformation observation data processing,by modeling and forecasting GPS observation adjustment data of Geheyan Dam at Qingjiang River before flood season between May and June in 1998.All the algorithms expounded is not related to non-linear estimation in this paper.