Scalar-Function Causal Discovery for Generating Causal Hypotheses with Observational Wearable Device Data.

Area of Science:

• Digital Health
• Causal Inference
• Health Data Analytics

Background:

• Wearable devices provide continuous, high-resolution functional data for health metrics.
• Traditional methods often focus on association, limiting causal mechanism discovery.

Purpose of the Study:

• To develop a new approach for generating causal hypotheses between continuous functional variables and binary scalar variables.
• To move beyond association-focused methods to reveal underlying causal mechanisms.

Main Methods:

• Proposed a scalar-function causal model identifiable with observational data.
• Developed a principled algorithm comparing likelihood functions of competing causal hypotheses.

Main Results:

• Theoretically demonstrated identifiability of the scalar-function causal model using observational data.
• Validated the method's robustness and applicability through simulations and real-world data.

Conclusions:

• The new approach enables causal hypothesis generation from continuous functional and binary scalar health data.
• The method is applicable to observational data, including wearable device data, for uncovering causal relationships.