Causal Algebras on Chain Event Graphs with Informed Missingness for System Failure.

Area of Science:

• Causal inference
• System reliability engineering
• Graphical models

Background:

• Bayesian networks (BNs) are commonly used for causal analysis in system reliability.
• BNs have limitations in fully expressing certain causal constructions relevant to reliability.
• Chain Event Graphs (CEGs) offer a more flexible alternative for capturing causal reasoning.

Purpose of the Study:

• To extend the use of CEGs for causal modeling in system reliability.
• To demonstrate CEGs' ability to define both remedial and routine maintenance interventions.
• To show CEGs' effectiveness in addressing causal inference complexities with missing data.

Main Methods:

• Utilizing Chain Event Graphs (CEGs) as an alternative graphical model to Bayesian networks.
• Developing a causal calculus for remedial interventions on CEGs.
• Extending the framework to incorporate routine maintenance interventions.
• Applying CEG methodology to address causal inference with missing data.

Main Results:

• CEGs effectively capture causal reasoning in system reliability, outperforming BNs for specific constructions.
• A causal calculus for remedial interventions was successfully devised on CEGs.
• CEGs can define both remedial and routine maintenance interventions.
• CEG methodology elegantly addresses causal inference complexities in missing data scenarios.

Conclusions:

• CEGs provide a powerful and flexible framework for causal inference in system reliability.
• CEGs offer advantages over BNs in modeling complex causal relationships and interventions.
• The CEG approach facilitates robust causal modeling and inference, even with incomplete data.