Abstract
Inference in nonlinear models is usually based on the asymptotic normal distribution, based on linearizing the model. The accuracy of this approximation can in many cases be improved by a reparametrization. Systematic methods for doing this will be described. Sometimes a saddlepoint approximation can be used, and this offers several advantages compared to the asymptotic distribution and the Edgeworth expansion. The improved methods are unfortunately not commonly used. It will be discussed why this is so. The methods will be illustrated by a series of examples.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Amari S-I (1982) Differential geometry of curved exponential families — curvatures and information loss. Ann Statist 10:357–385
Amari S-I (1985) Differential-geometrical methods in statistics. Lecture Notes in Statistics 28. Springer-Verlag
Bates DM, Watts DG (1980) Relative curvature measures of nonlinearity. JR Statist Soc B42:1–25
Bartlett MS (1936) The square root transformation in analysis of variance. JR Statist Soc Suppl 3:68–78
Box MJ (1971) Bias in nonlinear estimation. JR Statist Soc B33:171–201
Cook RD, Witmer JA (1985) A note on parameter-effects curvature. J Am Statist Assoc 80:872–878
Fieller EC (1954) Some problems in interval estimation. JR Statist Soc B16:175–185
Hinkley DV (1969) On the ratio of two correlated normal random variables. Biometrika 56:635–639
Hougaard P (1982) Parametrizations of non-linear models. JR Statist Soc B44:244–252
Hougaard P (1985a) Saddlepoint approximations for curved exponential families. Statist Prob Letters 3:161–166
Hougaard P (1985b) The appropriateness of the asymptotic distribution in a nonlinear regression model in relation to curvature. JR Statist Soc B47:103–114
Hougaard P (1986a) Covariance stabilizing transformations in nonlinear regression. Scand J Statist 13:207–210
Hougaard P (1986b) Survival models for heterogeneous populations derived from stable distributions. Biometrika 73:387–96 (Correction 75: 395)
Hougaard P (1988) The asymptotic distribution of nonlinear regression parameter estimates: Improving the approximation. Int Statist Rev 56:221–233
Pázman A (1984) Probability distribution of the multivariate nonlinear least squares estimates. Kybernetika 20:209–230
Pázman A (1990) Almost exact distributions of estimators II — Flat nonlinear regression models. Statistics 21:21–33
Pázman A (1991) Pivotal variables and confidence regions in flat nonlinear regression models with unknown σ. Statistics 22:177–189
Pázman A (1993a) Higher dimensional nonlinear regression — A statistical use of the Riemannian curvature tensor. Statistics 25:17–25
Pázman A (1993b) Nonlinear statistical models. Kluwer Academic Publishers
Skovgaard IM (1985a) Large deviation approximations for maximum likelihood estimators. Prob Math Statist 6:89–107
Skovgaard IM (1985b) A second-order investigation of asymptotic ancillarity. Ann Statist 13:534–551
Sprott DA (1973) Normal likelihoods and their relation to large sample theory of estimation. Biometrika 60:457–465
Væth M (1985) On the use of Wald's test in exponential families. Int Statist Rev 53:199–214
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Hougaard, P. Nonlinear regression and curved exponential families. Improvement of the approximation to the asymptotic distribution. Metrika 42, 191–202 (1995). https://doi.org/10.1007/BF01894299
Issue Date:
DOI: https://doi.org/10.1007/BF01894299