Causal inference and machine learning: we study how modern machine learning methods can be utilised for causal discovery and inference based on very large databases. The challenges we tackled are theoretical (identification theory, high dimensionality problems, asymptotic theory for inference) and methodological (algorithm capable of handling very large datasets in reasonnable computer time; secure computing solutions to be able to handle highly sensitive data).
Missing data and other selection mechanisms: combining registers, surveys and other longitudinal studies yields complex micro-data sets. Studies based on such data must carefully model sample selection and other missing data mechanisms. We develop such models and proposed methods to study deviations from the assumptions made (sensitivity analyses).
Health care performance: Open comparisons of hospitals and regions (benchmarking) provide information used in treatment priorities, patients’ choice of health care provider, and financial funding. The national quality registers allow for such comparisons and relevant statistical methods are developped to exploit this data and provide better support to quality improvement in health care.
Life course studies: Life course studies is a lively field of study in the social and health sciences, where the purpose is to study individual life trajectories. Life trajectories with different social and health domains can nowadays been observed and studied thanks to the availability of rich and large source of micro-data.
Free software: user friendly R functions are developed, making accessible the methods developed to the wider scientific community (see Software page).