29 July 2020 | 15:00 UTC/GMT
Time Zone Converter
The focus of biomedical and health research continues to shift towards large observational studies often using registries, population-based administrative databases and biobanks. The analysis of such data imposes many challenges and requires up-to-date knowledge of statistical methodology. There is, however, a huge gap between the rapid developments in cutting-edge methodology proposed in biostatistical literature and the statistical methods actually used in practice. Moreover, most methodological papers address a single analytical challenge, whereas in a typical real-life study many issues arise together.
To address these challenges, the STRATOS (STRengthening Analytical Thinking for Observational Studies) initiative was formed in 2013. This is a large, international, collaboration of experts in many different areas of biostatistical research. The aim is to systematically evaluate existing methodologies, identify unresolved issues, stimulate research in these areas and develop and disseminate guidance for practical analyses targeting researchers with different level of statistical expertise.
STRATOS currently has nine topic groups (TGs) and eleven cross-cutting panels, created to coordinate the activities of different TGs, share best research practices and disseminate research tools and results across TGs.
In this session we will present an overview of current developments and issues in four topic areas.
Willi Sauerbrei, Chair, STRATOS initiative, Institute of Medical Biometry and Statistic, University Medical Center Freiburg, Freiburg, Germany
Aris Perperoglou, AstraZeneca, Cambridge, UK
Towards state of the art for variable selection
Els Goetghebeur, Center for Statistics, Gent, Belgium
Three causal lessons from our Simulation Learner: SUTVA is fiction; radomisation fails as an instrument for many post randomisation exposures, and averaging causal effects over an observed (experimental) instrument may be irrelevant
James Carpenter, London School of Hygiene and Tropical Medicine, London, UK
Missing data: best practice and beyond in flexible modelling and causal inference
Michal Abrahamowicz, Department of Epidemiology and Biostatistics McGill University, Montreal, Canada
Designing simulation studies for accurate, generalizable conclusions: recent development