Abstract:Aiming at the low computational efficiency of the existing vector-parameter method for multivariable observations modeling and adjustment, we propose a matrix-parameter weighted least-squares method for multivariable observations, using the equivalent transformation, Kronecker product techniques and weighted least-squares method. Considering cross-correlation of different dimensional observations, the proposed method reconstructs the vector-parameter as a matrix-parameter form that has parameter independence and isomorphism in the multivariable observations model. This reduces the dimension number of the normal matrix and increases matrix density of the coefficient matrix. Consequently, high computational efficiency can be achieved with the proposed method. The adjustment results of space linear model and IGS coordinate time series model show that both the new matrix-parameter method and the traditional vector-parameter method have the same estimation results, but the proposed method achieves higher computational efficiency.