zbMATH — the first resource for mathematics

Generalized additive models: some applications. (English) Zbl 0633.62067
Generalized additive models combine the features of (parametric) generalized linear models and (nonparametric) additive models, which are a special case of the projection pursuit models of J. H. Friedman and W. Stuetzle, J. Am. Stat. Assoc. 76, 817-823 (1981): the linear parametric predictor in generalized linear models is replaced by a nonparametric additive model \(\eta (x)=\alpha +\sum f_ i(x_ i)\). The local scoring algorithm is used to estimate the functions \(f_ i(x_ i)\), using a scatterplot smoother as a building block.
The procedure may be used as a diagnostic tool for identifying parametric transformations of the covariates in a standard linear analysis. The link function, however, is assumed to be known.
Reviewer: L.Fahrmeier

62J99 Linear inference, regression
62G99 Nonparametric inference
Full Text: DOI