Simplified Confidence Intervals for an Awkward Nonlinear Model

摘要

A simple method is developed for approximating confidence intervals for a linear change-point model representing residential energy consumption. The method can be applied to any nonlinear model which is conditionally linear in all but one parameter. The approximation is shown to be more accurate than the Gaussian approximation, and easier to compute than confidence intervals defined by likelihood (sum-of-squares) regions. Computations are described in some detail for the motivating model. (Author)