Hello ALL, I just started to use segmented package for some piecewise linear fitting problem. However, author noted: "The package is not concerned with estimation of the number of the breakpoints. Although the BIC has been suggested, in general nonstatistical issues related to the understanding of the mechanism of the phenomenon in study could help to discriminate among several competing models with a different number of joinpoints.". Thus, judgement should be "theoretically" motivated. That is fine, and more than justified. But, what if… what would be the best way to "formally" decide about number of pieces? Many thanks, Petar [[alternative HTML version deleted]]