Parameter bounds for linear models in the presence of bounded relative errors

Parameter bounding in linear static models with bounded relative output errors is outlined, based on equation-error and errors-invariables models. Two equation-error procedures are proposed, pointing out approximated and over evaluation of bounds on the equation error. It is shown that the problem can be reformulated in terms of errors-in-variables models and a method has been outlined, based on previous results of the author (1990). Orthotope outer-bounding approximations of the feasible parameter region, which allow the evaluation of parameter uncertainty intervals, have been considered. Both the over-bounded equation-error (OBEE) and the bounded errors-in-variables (BEIV) approach have been tested on simulated data, pointing out the superiority of BEIV to OBEE as far as parameter uncertainty intervals and computing time are concerned