On modified information matrices in nonlinear regression

Rok, strany: 1996, 215 - 221

The Fisher information matrix does not depend on the curvature of a nonlinear regression model. Therefore, other information matrices are proposed, which are derived from two different origins: (a) from the probability density of the maximum likelihood, resp. of the maximum posteriori estimators, (b) from the second order derivative of the posterior density with respect to the Rao distance in the model.

ISO 690:

Pázman, A. 1996. On modified information matrices in nonlinear regression. In Tatra Mountains Mathematical Publications, vol. 7, no.1, pp. 215-221. 1210-3195.

APA:

Pázman, A. (1996). On modified information matrices in nonlinear regression. Tatra Mountains Mathematical Publications, 7(1), 215-221. 1210-3195.