Predictive and Specialized Modeling > Nonlinear Regression
Publication date: 04/12/2021

Nonlinear Regression

Fit Custom Nonlinear Models to Your Data

The Nonlinear platform is a good choice for models that are nonlinear in the parameters. This chapter focuses on custom nonlinear models, which include a model formula and parameters to be estimated. Use the default least squares loss function or a custom loss function to fit models. The platform minimizes the sum of the loss function across the observations.

Figure 15.1 Example of a Custom Nonlinear FitĀ 

Example of a Custom Nonlinear Fit

The Nonlinear platform also provides predefined models, such as polynomial, logistic, Gompertz, exponential, peak, and pharmacokinetic models. See Fit Curve.

Note: Some models are linear in the parameters (for example, a quadratic or other polynomial) or can be transformed to be such (for example, when you use a log transformation of x). The Fit Model or Fit Y by X platforms are more appropriate in these situations. For more information about these platforms, see Model Specification in Fitting Linear Models and Introduction to Fit Y by X in Basic Analysis.


Example of Fitting a Custom Model

Launch the Nonlinear Platform

The Nonlinear Fit Report

Nonlinear Platform Options

Create a Formula Using the Model Library

Additional Examples

Example of Maximum Likelihood: Logistic Regression
Example of a Probit Model with Binomial Errors: Numerical Derivatives
Example of a Poisson Loss Function
Example of Setting Parameter Limits

Statistical Details for the Nonlinear Platform

Profile Likelihood Confidence Limits
How Custom Loss Functions Work
Notes Concerning Derivatives
Notes on Effective Nonlinear Modeling
Want more information? Have questions? Get answers in the JMP User Community (