On the complexity of IQML algorithms

The authors study the computational complexity of two methods for solving least squares and maximum likelihood modal analysis problems. In particular, they consider the Steiglitz-McBride and iterative quadratic maximum likelihood (IQML) algorithms. J.H. McClellan and D. Lee (ibid., vol.39, no.2, p.509-12, 1991) have shown the iterations of the two methods to be equivalent. However, they suggest that the Steiglitz-McBride algorithm may be computationally preferable. A method for reducing the dimension of the matrix inversion required at each iteration of IQML is provided. The resulting reduction in the computation makes the computational complexity of IQML commensurate with that of the Steiglitz-McBride algorithm.<
>