Foundational Methods

Welcome to the foundational Methods pillar. This pillar develops advanced methods and tools for clinical prediction on observational data.

Mission

To establish and continuously refine the core methodologies, standards, tools, and infrastructure for utilizing observational healthcare data in actionable clinical prediction.

Research Question

What is the utility of observational healthcare data in clinical prediction, and which analytical methods best bring out its value?

Guiding Principles

Open science and correctness

Commit to transparency by openly sharing data, code, and methodologies. Maintain rigorous scientific standards and apply thorough testing from unit to integration and regression tests to ensure models are correct, reliable, and useful to the broader research community.

Performance

Prioritize computational efficiency and rapid training cycles by designing algorithms and systems that deliver accurate predictions quickly and at scale.

Reproducible

Ensure that all research is fully reproducible through detailed documentation, standardized protocols, and strict version control so that experiments and results can be independently verified and extended.

Reusable

Develop modular algorithms, code, and tools that facilitate easy reuse and integration into other projects, thereby supporting efficient collaboration and the ongoing evolution of our research.