We can use this value as a starting point to calculate g, l, and S,
and then update b. The procedure can be repeated until convergence.
Since b0 and bL can be obtained using Ordinary Least Squares
(OLS), the iterative ML method takes advantages of OLS and can be
easily implemented. Fig. 1 shows the whole procedure of the iterative
ML method. In this algorithm, the optimization method
employed to optimize the concentrated likelihood Lc is the BFGS
method. For the spatial SUR with spatially autocorrelated error, it is
not necessary to calculate bL. A Matlab routine has been developed
for the iterative ML method and is available for requesting