Im Rahmen des Biometrischen Kolloquiums 2021 werden drei Tutorien angeboten.
|“Introduction to Machine learning with R“||halbtägig, vormittags||Przemyslaw Biecek|
|“rpact R Package for Adaptive Confirmatory Trials”||halbtägig, nachmittags||Gernot Wassmer|
|“An Introduction to Causal Inference and Target Trials”||ganztägig||Sonja Swanson|
Die Tutorien finden in englischer Sprache am Sonntag dem 14.03.2021 statt.
Die weiteren Informationen sind hier daher auch in Englisch gehalten.
Tutorial “Introduction to Machine Learning with R” (half-day, morning), Przemyslaw Biecek
– The two cultures – To predict or to understand
– Evaluation of black-box models (RMSE, AUC, ROC, F1, LIFT, train-test split, CV-folds)
– Towards complex models – bias variance tradeoff (random forest, bagging, boosting)
– Explanatory Model Analysis – global level (partial dependence profiles, feature importance)
– Explanatory Model Analysis – local level (Shapley values, ICE/CP)
– The multiplicity of good models – Rashomon effect
– From complex to interpretable – how to extract features from black-box models
With a hands-on part based on R examples with medical examples.
– Introduction to Statistical Learning: http://faculty.marshall.usc.edu/gareth-james/ISL/
– Explanatory Model Analysis: https://github.com/pbiecek/ema
– mlr3 book: https://mlr3book.mlr-org.com/
Przemyslaw Biecek graduated in Mathematical Statistics and Software Engineering from Wroclaw University of Technology. He obtained his PhD in biostatistics, after which he held positions at Hasselt University (Belgium), Davis University (USA) and Nanyang Technological University (Singapore). He is now an Associate Professor in Machine Learning at Warsaw University of Technology and University of Warsaw.
His main areas of research include model visualization, model interpretability and predictive modelling of large and complex data with a special focus on applications in personalized computational medicine/oncology. He is also interested in evidence-based medicine and statistical software engineering (an R enthusiast).
Tutorium “rpact R Package for Adaptive Confirmatory Trials” (half-day), Gernot Wassmer
– Concept of the Software
– Methodology and Basis Features of the Software
– Shiny App / Vignettes
– Designing Adaptive Trials
Gernot Wassmer has a diploma in Statistics and received his PhD from the University of Munich, Germany in 1993, after which he was a Research Fellow at Munich’s Institute of Statistics, at the Institute for Epidemiology, GSF Neuherberg, and at the Institute of Medical Statistics, University of Cologne, Germany. From 2010 to 2014, he worked at Aptiv Solutions and Icon in Cologne. He is an Adjunct Professor of Biostatistics at the Institute of Medical Statistics, University of Cologne, and founder and owner of RPACT GbR. His major research interest is in the field of statistical procedures for group sequential and adaptive plans in clinical trials. He has been a member of independent data monitoring committees for international, multi-center trials in various therapeutic fields and also serves as a consultant for the pharmaceutical industry.
Tutorium “An Introduction to Causal Inference and Target Trials” (full-day), Sonja Swanson
– Basic concepts and assumptions of causal inference, using counterfactual or potential outcomes
– Key sources of bias: e.g., confounding, selection, and information bias
– Describing a target trial: describing the key protocol elements of an ideal randomized trial, including the eligibility criteria, treatment strategies, treatment assignment, follow-up period, outcome, causal contrast, and statistical analysis
– Emulating a target trial: designing and analyzing observational data to estimate causal effects, including the use of g-methods
Cain LE, Saag MS, Petersen M, May MT, Ingle SM, Logan R, et al. Using observational data to emulate a randomized trial of dynamic treatment-switching strategies: an application to antiretroviral therapy. Int J Epidemiol. 2016; 45(6):2038–49.
García-Albéniz X, Hsu J, Hernán MA. The value of explicitly emulating a target trial when using real world evidence: an application to colorectal cancer screening. Eur J Epidemiol. 2017. doi:10.1007/s10654-017-0287-2.
Hernán MA, Robins JM. Using big data to emulate a target trial when a randomized trial is not available. Am J Epidemiol. 2016; 183(8):758–64.
Labrecque, J.A., Swanson, S.A. Target trial emulation: teaching epidemiology and beyond. Eur J Epidemiol. 2017; 32, 473–475.
Sonja Swanson is an Assistant Professor in the Department of Epidemiology at Erasmus MC and holds an adjunct affiliation with the Department of Epidemiology at the Harvard T. H. Chan School of Public Health. She was recently invited to join the editorial team of Epidemiology. Her methodological research focuses on improving the use and transparency of methods for estimating causal effects in epidemiology. This work spans applications in observational studies and pragmatic randomized trials. She has made important contributions to instrumental variables methods, e.g. in the context of Mendelian randomization, as well as target trials, effect heterogeneity or causal mediation. Her substantive research primarily focuses on neuropsychiatric disorders and related health outcomes, including using appropriate methods to studying potential prevention and treatment strategies.