Confidence Investigation of Discovering Organizational Network Structures Using Transfer Entropy

Transfer entropy has long been used to discover network structures and relationships based on the behavior of nodes in the system, especially for complex adaptive systems. Using the fact that organizations often behave as complex adaptive systems, transfer entropy can be applied to discover the relationships and structure within an organizational network. The organizational structures are built using a model developed by Dodd, Watts, et al, and a simulation method for complex adaptive supply networks is used to create node behavior data. The false positive rate and true positive rates are established for various organizational structures and compared to a basic tree. This study provides a baseline understanding for the accuracy that can be expected when discovering organizational networks using these techniques. It also highlights conditions in which it may be more difficult to successfully discover a network structure using transfer entropy and bounds confidence levels for practitioners of such methods.