Model-based recursive partitioning of structural equation models

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] The goal of this dissertation is to describe the algorithmic method of model-based recursive partitioning, and to compare different estimation techniques when the algorithm is applied to structural equation models. Model-based recursive partitioning (MOB) is an extended version of decision trees that can handle stochastic models. Thus, the method is intensive and involves estimation of stochastic models in every recursion (i.e., in every tree branch). We compare the two estimation methods of likelihood ratio test and generalized M-fluctuation test. The tests' background and implementation are detailed, and the tests' abilities to identify the population models are studied via simulation. A real-world example involving the tests is also provided, followed by a discussion of extensions and future development.