Causal Loop Diagram Aggregation Towards Model Completeness
Experienced system dynamicists commonly conceptualise causal relationships and feedback loops using Causal Loop Diagrams (CLDs). In adhering to best practice, multiple data collection activities may be required (e.g. multiple group model building sessions), resulting in multiple CLDs. To achieve covariation that correctly attributes cause and effect from multiple data sets, aggregation of CLDs may be necessary. Such aggregation must adequately account for attribution variations across constructed CLDs to produce a coherent view of a phenomenon of interest in a 'complete' model. Discourse concerning model completeness should account for the potential for method bias. The data collection method chosen for CLD development will influence the ability to create a model that is fit for the purposes of the study and influence the likelihood of achieving model completeness. So too does the method chosen for model aggregation. Little processual guidance exists on a method for data aggregation in system dynamics studies. This paper examines three data aggregation approaches, based on existing qualitative analysis methods, to determine the suitability of each method. The approaches considered include triangulation, includes all data in the aggregation process; grounded theory, bases aggregation on frequency of occurrence; and synthesis, extends aggregation to include variables based on magnitude of occurrence. Comments are made regarding the relevance of each method for different study types, with final remarks reiterating the consideration of equifinality and multifinality in research and their impact on method selection. This paper enhances the rigour of research aiming at facilitating greater success in studies utilising CLDs.