Linear regression pathmox segmentation tree : the case of visitors' satisfaction to attend a Spanish football match at the stadium

Segmentation trees have been attracting a great deal of attention as model comparison tools, with research mainly motivated by the fact that segmentation trees allow identification of partitions of data characterised by different dependency structures. Few algorithms have been proposed by the statistical community that combine model estimation and segmentation trees, outside the MOdel-based recursive partitioning (MOB) procedure proposed by Zelies et al. (2008). In a new approach we generalize the pathmox algorithm developed by Lamberti et al. (2016) to the context of linear regression models, using a model comparison test to identify the most significant partitions (i.e., sub-groups) in data. Further developments of the proposed approach will involve extensions to other contexts such as quantile regression.