Nonlinear Explanatory Multiple Regression Models

Abstract

In this chapter, the insights offered by nonlinear regression models built with multiple drivers are examined. In many cases, a dependent variable and its drivers are approximately Normal, with skewness between −1 and +1. Linear regression models often provide good fit for either cross sectional or time series, and are often valid for forecasting in time series. However, the choice of a nonlinear model enables acknowledgement of the interactions inherent in many cases. With nonlinear models, the impact of each driver depends on the values of other drivers. In a nonlinear model, driver influences are multiplicative, which adds an element of realism relative to linear regression models with constant response, and provides richer insights from sensitivity analysis.