Innovative statistical and machine learning methods for comparing performance and outcome in register data studies

Statistical challenges within different applications are often similar.

The overall aims of this interdisciplinary project are to develop, evaluate, and implement innovative statistical methods, including modern machine learning methods, to be used with register data, that share similar features.

Although the proposed methods will be evaluated with data from educational and health registers, the methods are generic and can be used with other large-scale registers.

The research has three themes: 1) Develop innovative statistical methods that combine information from different measurement scales and instruments and improve score norming.

A challenge when combining different register information is when different instruments, questions and scales have been used at different time points.

The questions may also be of different difficulty, content or used in different contexts.2) Develop and adapt modern machine learning methods to examine inequalities in outcome and performance using register data.

This facilitates fair and stable comparisons.3) Implement the new methods with register data and develop freely available software for researchers and practitioners.The methods will also be evaluated in close collaboration with researchers from the educational and health sciences.

The project will have a large impact on register-based research and be valuable for policymakers, researchers, and those who work with large-scale registers.