Evidence of effectiveness.

Area of Science:

• Epistemology of science
• Causal inference
• Philosophy of statistics

Background:

• Two views exist on intervention effectiveness: reliance solely on population studies vs. inclusion of mechanistic evidence.
• Assessing these views requires a general normative principle of inference.

Purpose of the Study:

• To evaluate arguments for and against using mechanistic knowledge in causal inferences about interventions.
• To assess the role of mechanistic evidence within the Bayesian framework.

Main Methods:

• Analysis of competing views on intervention effectiveness.
• Application of Bayesian inference principles (likelihood and prior probabilities).
• Examination of how mechanistic knowledge influences prior probabilities.

Main Results:

• Bayesian inference naturally incorporates both statistical (likelihood) and mechanistic (prior) evidence.
• Mechanistic knowledge is a valid influence on the prior probability of causal hypotheses.
• The Bayesian framework supports integrating mechanistic evidence for causal claims.

Conclusions:

• Causal inferences about intervention effectiveness benefit from the integration of mechanistic evidence.
• The Bayesian approach provides a theoretical basis for combining statistical and mechanistic data.
• This study supports the view that mechanistic knowledge is crucial for robust causal inference.