Abstract
Statistical models commonly assume that the relation between a predictor and a criterion can be described by a straight line. This assumption is often appropriate, but there are times when abandoning it is warranted. Under these circumstances, we have two choices: adapt a linear model to accommodate nonlinear relations (e.g., transform the variables; add cross product terms) or use statistical techniques that directly model nonlinear relations.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Other approaches to smoothing nonlinear relations, such as kernel regression and locally weighted regression (LOESS), are discussed in Brown (2014).
- 2.
The data in Fig. 9.3 are fictitious, but the effect is all too real!
- 3.
Natural splines are defined only for odd degrees, and natural cubic splines are most common.
- 4.
Penalized smoothing splines are also called penalized regression splines.
- 5.
The edf is sometimes referred to as the Effective Degrees of Freedom
- 6.
The \( \mathrm{\mathcal{R}} \) syntax at the end of this section describes the operations that produced the knot sequence.
- 7.
When implementing P-splines, Eilers and Marx use divided differences to create an equivalent B-spline basis.
- 8.
Cross-validation and generalized cross-validation usually produce very similar fits (Craven & Wahba, 1979).
- 9.
The covariance matrix described in Eq. (9.24) is known as the Bayesian posterior covariance matrix.
- 10.
Additive models can be also used with distributions that are not Gaussian. These Generalized Additive Models, as they are known, will be discussed in Chap. 11.
- 11.
- 12.
If you compare the fitted values using penalized least squares from the ones obtained with backfitting, you will see that the estimates differ slightly but are substantially correlated (r = .9991). This will ordinarily be the case, so the choice of which \( \mathrm{\mathcal{R}} \) package to use (gam or mgcv) is largely personal.
References
Brown, J. D. (2014). Linear models in matrix form: A hands-on approach for the behavioral sciences. New York: Springer.
Breiman, L., & Friedman, J. H. (1985). Estimating optimal transformations for multiple regression and correlations (with discussion). Journal of the American Statistical Association, 80(391):580–619. https://doi.org/10.2307/2288473. JSTOR 2288473.
Craven, P., & Wahba, G. (1979). Smoothing noisy data with spline functions. Numerische Mathematik, 31, 377–403.
Eilers, P. H. C., & Marx, B. D. (1996). Flexible smoothing with B-splines and penalized likelihood (with discussion). Statistical Science, 11, 89–121.
Eilers, P. H. C., & Marx, B. D. (2010). Splines, knots, and penalties. Computational Statistics, 2, 637–653.
Hastie, T. J., & Tibshirani, R. (1986). Generalized additive models. Statistical Science, 1, 297–318.
Hastie, T. J., & Tibshirani, R. (1990). Generalized additive models. London: Chapman and Hall.
Reinsch, C. H. (1967). Smoothing by spline functions. Numerische Mathematik, 10, 177–183.
Wood, S. N. (2003). Generalized additive models: An introduction with R. London: Taylor & Francis.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer International Publishing AG, part of Springer Nature
About this chapter
Cite this chapter
Brown, J.D. (2018). Cubic Splines and Additive Models. In: Advanced Statistics for the Behavioral Sciences. Springer, Cham. https://doi.org/10.1007/978-3-319-93549-2_9
Download citation
DOI: https://doi.org/10.1007/978-3-319-93549-2_9
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-93547-8
Online ISBN: 978-3-319-93549-2
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)