UC San Diego

Causal inference plays a pivotal role in statistical analysis and decision-making across various disciplines, including epidemiology, economics, and social sciences. Database systems, as the backbone of information storage and retrieval across multiple sectors, require sophisticated analytical capabilities to support decision-making processes. However, real-world datasets often present complexities such as redundancy, incompleteness, and the lack of critical attributes. This thesis proposes multifaceted approaches to address these complexities and limitations of traditional causal inference methods in database environments. Specifically, it introduces a graphical user interface that leverages what-if and how-to queries. Additionally, we create a framework that enriches datasets from diverse data sources to uncover complex relationships, ensuring robust causal analysis.