The book is devoted to nonlinear regression models, which is to be linearized. The problem is whether neglected terms in the Taylor series do not cause an essential deterioration of results of statistical solutions. Therefore the main attention is devoted to a construction of criteria which enable us to decide whether the linearization is admissible. These criteria are found for an estimation, konfidence ellipoids and testing statistical hypotheses. Model without constraints on paramaters, models with the type I constraints and models with the type II constaraints are studied.