Causal driver detection with deviance information criterion
Shing H. Doong, Tean Q. Lee · 2010
Causal explanatory study is a very important research method in empirical research. The outcome of a quantitative MIS research frequently reports significant factors of a causal model. Locating causal drivers is in some sense similar to feature selection in data mining. This study uses Bayesian regressions and Markov Chain Monte Carlo simulations to detect drivers in a research model of information systems study. Deviance information criterion is used to compare Bayesian models resulted from different prior distributions. Differential evolution and a deterministic type iterative procedure are proposed to find the best prior distribution, which is used to find drivers of the final Bayesian regression model. Experimental results show that these approaches can locate more interesting drivers of the research model.