Bias of the M. L. E. in singular nonlinear regression models

Rok, strany: 1999, 55 - 63

In a singular nonlinear regression model with normal and independent errors we derive expressions for the approximation of the bias of the maximum likelihood estimators of estimable parameter functions. So we extend results known in regular models to singular ones. This is done by putting equality constrains on the parameters, to make all parameters identifiable and by approximating the normal equations of the estimator up to $o_p(N^-1)$. Invariance of the approximate bias to reparametrizations and to the choice of identifying constraints is also proved.

ISO 690:

Pázman, A. 1999. Bias of the M. L. E. in singular nonlinear regression models. In Tatra Mountains Mathematical Publications, vol. 17, no.3, pp. 55-63. 1210-3195.

APA:

Pázman, A. (1999). Bias of the M. L. E. in singular nonlinear regression models. Tatra Mountains Mathematical Publications, 17(3), 55-63. 1210-3195.