Application of big data Bayesian interrupted time-series modeling for intervention analysis

Recent technological advancements have further shifted the focus in data collection to now allow temporal data analysis techniques to proliferate. However, while time-ordered sequences of observations provide opportunities to draw richer insights, time series methodologies in social research are still scarce. Therefore, to address these growing methodological needs, this book chapter introduces the use of interrupted time series (ITS) design for intervention analysis to examine and understand the dynamics of social interventions over time. Specifically, this book chapter aims at drawing attention to a novel decision science-based intervention analysis technique (i.e., Bayesian Interrupted Time-Series Analysis), while performing an in-depth review of the time series literature related to intervention analysis. Bayesian models are apt to estimate the impact of measures, policies, and interventions as they have been found to provide reliable outcomes even with correlated data estimates, and to allow model parameters to evolve over time while including prior parameter information as well. By doing so, we also assert the significance of the use of such intervention analysis method for event studies over dynamic time series. This book chapter will serve as an integrative resource for social science researchers interested in using temporal data to assess other social interventions, in observational studies such as to assess the impact of a wide range of policies, measures and interventions, which all require the assessment of causality over time.