A new method has been presented for the determination of a low-order model approximating a high-order system. It is based on the use of the matrix pseudo-inverse to estimate the parameters of the model which minimize the sum of the squares of the errors between the response of the actual system and that of the model at the sampling instants. One of the advantages of this method is that the description of the actual system dynamics need not be known, but only the measurements of the input-output data are required. As the algorithms are iterative, computation is fairly straightforward, and the requirement for storage of input-output data depends only on the order of the assumed model, not on the number of iterations in the interval considered for minimization. An example of the application of the method for the determination of an approximate second-order model of a seventh-order system has been given, and compared with the reduced model obtained using another method.