Causal models and learning from data: integrating causal modeling and statistical estimation.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Epidemiology relies on causal questions to understand disease.
• Formal causal inference frameworks offer potential for increased rigor.
• The role of formal causal thinking in applied epidemiology is debated.

Purpose of the Study:

• To advocate for the utility of formal causal thinking in epidemiology.
• To clarify the capabilities and limitations of causal models.
• To provide an accessible introduction to causal modeling tools.

Main Methods:

• Presenting a systematic approach to integrate causal modeling with statistical estimation.
• Highlighting common challenges in applying causal modeling techniques.
• Discussing the types of questions addressable by causal frameworks.

Main Results:

• Formal causal frameworks aid in designing statistical analyses that approximate causal questions.
• These frameworks clarify the assumptions needed for causal interpretation of estimates.
• A structured method for combining causal modeling and statistical estimation is outlined.

Conclusions:

• Formal causal thinking is valuable for improving the rigor of epidemiological research.
• Causal models, when properly applied, enhance the ability to answer causal questions.
• Understanding causal models and their assumptions is crucial for applied epidemiologists.