Ontological Models Supporting Covariates Selection in Observational Studies.

Area of Science:

• Biostatistics and Causal Inference
• Knowledge Representation and Ontologies
• Data Modeling and Analysis

Background:

• Biostatisticians utilize causal diagrams for covariate selection in multivariate modeling.
• Traditional causal diagrams have limitations in representing complex relationships like interactions and bidirectionality.
• The MetBrAYN project addresses these limitations through an innovative ontological approach.

Purpose of the Study:

• To develop an ontological-based process for building enhanced causal diagrams.
• To integrate biostatisticians' methodological knowledge into a general ontology.
• To semi-automatically generate ontology-based causal diagrams for improved covariate selection.

Main Methods:

• Developing a general ontology to encapsulate biostatisticians' knowledge.
• Utilizing the ontology as a wrapper to aggregate diverse knowledge forms.
• Implementing inference rules to build and curate ontology-based causal diagrams.
• Semi-automating the construction of causal diagrams.

Main Results:

• A novel system for creating ontology-based causal diagrams has been established.
• The system facilitates the aggregation of various knowledge types within an ontology.
• Biostatisticians can curate and visualize recommended covariates for specific research questions.
• Limitations of traditional causal diagrams are addressed by the new approach.

Conclusions:

• The MetBrAYN project offers a robust ontological framework for causal inference.
• This system enhances the process of covariate selection for multivariate models.
• The approach supports biostatisticians in building more accurate and comprehensive causal diagrams.
• Future work may involve further refinement and application of the ontology-based system.