Abstract for: The Ford Method: A Sensitivity Analysis Approach

In dynamic models, a system's behavior is determined by the interaction of its feedback loops. The challenge for system dynamics modellers is to identify these loops, and also understand, over the runtime of a model, which loops dominate system behavior. The Ford method is a procedure that identifies changes in atomic behavior patterns in the presence, and absence, of feedback loops, in order to identify loop dominance for a specific time interval. For a candidate feedback loop, dominance is calculated based on setting the loop control variable to a constant value. Our approach proposes a variation on this method. Using sensitivity analysis, we explore a wider search space around a range of possible values for the control variable, and thereby the value of the loop gain. The outcome of this analysis is a richer set of loop dominance analysis for each atomic behavior pattern in the model. The sensitivity of various feedback loops is measured as an indicator of dominance. This approach has the potential to be selected for modellers and policy makers to analyze the structure-drive-behavior dynamic systems. The approach is illustrated through an analysis of the Yeast Model.