Savchenko A. Adopted T(q)-likelihood estimator in non-linear errors-in-variables models.

The thesis is devoted to T(q)-likelihood estimator in nonlinear structural errors-in-variables models, where measurement error is classical additive with known variance. Exponential family of densities for response variable is under consideration. The consistency and asymptotic normality conditions of the estimator for known and unknown dispersion parameter are established. Two cases of construction of T(q)-likelihood estimator in exponential family of densities with unknown dispersion parameter are under consideration. Milder consistency and asymptotic normality conditions are presented for particular models (i.e., for exponential, Poisson, linear, quadratic, and Gamma model).