Linearization of nonlinear regression models by smoothing

Rok, strany: 2001, 13 - 25

We suggest linear estimators of parameters of a nonlinear regression model, which are derived from a linearization “by smoothing”. We compare them with the estimators coming from a “standard” linearization (by the use of a Taylor formula). The proposed estimators has a better mean squared error (especially in the case that the regression is quadratic in the unknown parameters). The method can be applied also when the regression function has no derivatives, and in general, the necessary coefficients can be obtained by simulations. If the prior distribution of the parameters is much concentrated, the proposed and the standard methods coincide.

ISO 690:

Pázman, A. 2001. Linearization of nonlinear regression models by smoothing. In Tatra Mountains Mathematical Publications, vol. 22, no.2, pp. 13-25. 1210-3195.

APA:

Pázman, A. (2001). Linearization of nonlinear regression models by smoothing. Tatra Mountains Mathematical Publications, 22(2), 13-25. 1210-3195.