Statistical methods for model comparison in parameter estimation problems for distributed systems (Q808588)

The paper deals with nonlinear regression problems. The regression function is assumed to belong to a parametric family, typically generated as a solution of a parametric partial differential equation. For statistical analysis, the least squares approach is adopted and the usual limit theorems for least squares estimators are proved. The exogeneous variables are assumed to be generated by a known probability measure. The innovational point of the paper is that the parametric family of regression functions is not known exactly but only approximately. The validity of the limit theorems is proved under a simultaneous increase of the accuracy of the regression function.