Identification of a spatial autoregression by rank methods

For the process of a spatial autoregression of order (1, 1), we constructed locally the most powerful rank criteria for testing the hypotheses of coefficients of autoregressive equation. Statistics of criteria at zero hypothesis are free from distribution and are asymptotically normal. Basing on statistics of rank criteria, we proposed an algorithm for constructing point estimates of coefficients of autoregressive equation. The designed methods for estimation and test of hypotheses are resistant to ‘outliers’ in observations.