Review article

The paper considers major disadvantages of the parametric theory of regression experiment design and analysis. Alternatively, non-parametric approaches to such problems are recommended. It is pointed out that non-parametricity in the stated problems occurs given poor a priori information on the regression model, under measuring infeasibility "at points" enforcing the conversion using measurement schemes at "functionals" or "operators", as well as when selecting space and estimation operators. The advantages of nonparametric formulations for regression experiment design and analysis along with the scope of their applicability are considered.