Uncovering Complex Relationships in System Dynamics Modeling: Exploring the Use of CART, CHAID and SEM

摘要

One of the premises of system dynamics is that the modeler would make relationship assumptions with enough precision to make the model useful. A common validation method is to consult with field experts, but with the advent of the internet, and automated data collection methods, knowledge is diluted as companies store abundant information without time to process it. Customers' dislikes, perceptions, intentions, opinions, and service characteristics reside in data warehouses (e.g. survey data is stored as categorical, nominal, ordinal or qualitative without further analysis). Without experts, companies are data rich but not necessarily knowledge rich. We present an application of known nonparametric predictive methodologies to uncover/confirm significant variable relationships and build the equations to feed the model: Classification and Regression Trees (CART), Chi-Square Automatic Interaction Detection (CHAID) and Structural Equation Modeling (SEM). A developing application of CHAID/SEM to explore restructuring decisions in a large service organization will be briefly discussed.