Piecewise regression: a simplified procedure.

A recurring problem is the estimation of regression parameters when the data sample is hypothesized to have been generated by more than a single regression model. This problem of ‘piecewise regression’ has been examined previously by using a combination of regression theory and hierarchical cluster analysis. The approach espoused here is applicable in a limited domain, but when it applies allows a solution to be found with very little calculation. Tukey's ‘jackknife’ is used to test the meaningfulness of any suggested subdivision of the regression curve into pieces. Two examples are included.