Detecting a change in the type of relationship modelled in regression form is accompanied by the need for inference about the point at which the change occurs. This research will examine statistical tests and their properties and will construct confidence regions for change points, concentrating on two-phase multiple regression. Approximate distributions for test statistics will be generated, power functions will be estimated, robustness will be considered theoretically and numerically, and an alternative estimator will be defined using a neural network algorithm.
