Statistical Methods for Causal Inference Research Group
This unit is concerned with statistical methods aiming at drawing reliable conclusions from observational data so as to assess the consequences of specific (sometimes hypothetical) interventions. Such insights are crucial for planning future intervention studies, and ultimately for public health decision making and policies, especially in the context of preventative measures and strategies. The particular statistical challenge consists in adequately addressing weaknesses and limitations of the data, such as lack of, or imperfect, randomization, systematic selection or drop-out etc. These need to be accounted for by suitable statistical models and methods.
The unit focusses on investigating and comparing the theoretical and practical properties of such methods, and aims at developing new or modifying existing methods especially relevant to the statistical analyses carried out at BIPS. A prerequisite is the in-depth understanding and scrutinizing of the underlying assumptions so that these can be made plausible, either empirically or based on subject matter knowledge.
The research interests of this unit are:
- Confounder selection and adjustment
- Models and methods for causal mediation analysis
- Causal models and methods for time-dependent data, especially time-varying exposures
- Instrumental variables, especially “Mendelian Randomisation”
- Bias modelling and analysis
- Vansteelandt S, Didelez V. Improving the robustness and efficiency of covariate-adjusted linear instrumental variable estimators. Scandinavian Journal of Statistics, Theory and Applications. 2018;45(4):941-961.
- Farewell D, Huang C, Didelez V. Ignorability for general longitudinal data. Biometrika. 2017;104(2):317-326.
- Jones E, Didelez V. Thinning a triangulation of a Bayesian network or undirected graph to create a minimal triangulation. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems. 2017;25(3):1750014.
- Didelez V, Evans R. Causal inference from case-control studies. In: Borgan O, Breslow N, Chatterjee N, Gail M, Scott A, Wild C, editors. Handbook of statistical methods for case-control studies. Boca Raton, Florida: Chapman & Hall/CRC. 2018. S. 87-115.
- Didelez V. Discussion on 'Causal inference by using invariant prediction: Identification and confidence intervals' by Peters, Bühlmann, Meinshausen. Journal of the Royal Statistical Society. Series B (Statistical Methodology). 2016;78(5):990-991.