Locating Event-Based Causal Effects: A Configural Perspective.

Area of Science:

• Statistics
• Causal Inference
• Evolutionary Biology

Background:

• Traditional statistical models for directional hypotheses rely on linear models, limiting the analysis of causal theories where effects depend on whether causes occurred.
• Existing methods do not fully accommodate theories defining effects based on the presence or absence of specific causes.

Purpose of the Study:

• To propose and specify statistical models for analyzing directional or causal hypotheses at the configural level, accommodating theories where causes influence effect densities.
• To introduce base models for the configural analysis of directional hypotheses using log-linear models of direction dependence in a confirmatory context.

Main Methods:

• Development of configural analysis models for directional hypotheses, analyzing how causes affect data space density.
• Application of log-linear models for direction dependence in a confirmatory framework.
• Simulation studies to assess the power of proposed methods in identifying density deviations.

Main Results:

• The proposed configural analysis methods demonstrate strong power in detecting sectors of the data space with expected or unexpected case densities.
• A data example confirms the evolutionary hypothesis that body size influences brain size, particularly in higher vertebrates.

Conclusions:

• Configural analysis provides a robust framework for testing directional and causal hypotheses, especially when effects are contingent on the occurrence of causes.
• The developed statistical models offer a powerful tool for confirmatory analysis of direction dependence and have empirical support in evolutionary biology.