23. Nonlinear Models
A key tenet of linear models is a linear relationship, which is actually reflected in the coefficients, not the predictors. While this is a nice simplifying assumption, in reality nonlinearity often holds. Fortunately, modern computing makes fitting nonlinear models not much more difficult than fitting linear models. Typical implementations are nonlinear least squares, splines, decision trees and random forests and generalized additive models (GAMs).
23.1 Nonlinear Least Squares
The nonlinear least squares model uses squared error loss to find the optimal parameters of a generic (nonlinear) function of the predictors.
Get R for Everyone: Advanced Analytics and Graphics, 2nd Edition now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
