A Simulation Study of Pathmox with non-normal Data

One of the current research topics in PLS-PM is the issue of segmentation, i.e. examining whether the analyzed population is heterogeneous. Two types of heterogeneity can be distinguished: observed and unobserved. In observed heterogeneity subpopulations can be defined based on observed variables. In unobserved heterogeneity the variables that cause heterogeneity are unknown. When dealing with observed heterogeneity in PLS-PM, one can find the PATHMOX approach. PATHMOX produces a segmentation tree in which each node has an associated path model. As a split criterion it employs an F-based statistic for testing the equality of path coefficients between two compared models. This paper discusses a simulation study to evaluate the sensitivity of the test when it is applied for comparing two inner path models under different experimental conditions using non-normal data.