STA517 Nonlinear Regression Modelling (8)

Most statistical procedures implicitly assume a linear model. This subject investigates the modelling of truly nonlinear data via applications. While investigation of the theory of nonlinear estimation is necessary, an empirical approach to demonstration of the technique is given emphasis by the use of computer packages using real data. Advanced topics such as the use of curvature measures are introduced.

No offerings have been identified for this subject in 2021.

Where differences exist between the Handbook and the SAL, the SAL should be taken as containing the correct subject offering details.

Subject Information

Grading System



One session


School of Science and Technology

Assumed Knowledge

Learning Outcomes

Upon successful completion of this subject, students should:
  • Recognise situations for which a nonlinear model is appropriate;
  • Understand the basic theory underlying the definition, role and applications of nonlinear regression models;
  • Appreciate the difficulties entailed in fitting nonlinear regression models to normal data;
  • Be able to emphasise the usefulness of the nonlinear approach using applications to practical situations;
  • Be able to perform analysis tasks on recognised software platforms (SPLUS) and interpret results in terms of the original problem;
  • Understand the theory and application of reparameterisation based on curvature measures.


This subject will cover the following topics:

Introduction The main features of the nonlinear regression model are highlighted with comparison to the generalized linear model and transformations. The use of a statistical software package (S-PLUS) for nonlinear regression models is also covered. Theory of Estimation This topic focuses on the mathematics underlying the nonlinear regression model. Normal equations, the linearization technique and the geometry of least squares for the nonlinear case are covered. Problems with search procedures are also highlighted. Fitting Procedures This topic introduces the different types of fitting procedures for nonlinear regression models including linearization, Newton Raphson, Steepest Descent and Marquardt's Correction. The problems associated with the choice of starting values are discussed. The use of curvature measures and the connection to model parameterization are a key aspect of the topic. Inference The use and practical application of curvature measures to assess nonlinearity are expounded with particular emphasis on statistical software packages such as S-PLUS. Model assessment and the assumptions underpinning the nonlinear regression model are discussed.

Residential School

This subject contains a 2 day Optional Residential School.Lecture/workshop

The information contained in the CSU Handbook was accurate at the date of publication: May 2021. The University reserves the right to vary the information at any time without notice.