2:30–3:30 pm
Jones 226 5747 S. Ellis Avenue
Tobias Friedling, University of Cambridge
"Selective Randomization Inference for Adaptive Studies"
Abstract
Many clinical trials follow a design with multiple stages: After each stage, the data is provisionally analysed and – based on these results – the recruitment of participants for the next stage as well as the administered treatment is chosen adaptively. For instance, we may want to exclude poorly performing drugs early (multi-arm multi-stage trials) or gather more samples from a certain subpopulation that shows a potentially beneficial response (enrichment trials).
Analysing such adaptive studies is challenging as the data is used twice: (1) for selection of the design of later stages and the null hypothesis, (2) for testing the null hypothesis with the data generated under the chosen design. Since the data generating mechanism and null hypothesis are not pre-specified, classical statistical methods do not provide valid inference. The literature on adaptive studies is aware of this issue; however, proposed solutions are limited in scope and usually specific to a certain design.
We propose a general and assumption-lean framework for analysing adaptive studies that combines concepts from randomization and post-selection inference. We show that our method improves power while still controlling the selective type-I error. Moreover, we address the construction of selective confidence intervals and the associated computational challenges.
This is an ongoing project joint with Zijun Gao and Qingyuan Zhao.
Bio:
Tobias is a 4th year PhD student at the Statistical Laboratory of the University of Cambridge supervised by Qingyuan Zhao. He completed his Bachelor and Master degree in Munich (thesis advisor: Mathias Drton) and mainly works on causal inference, sensitivity analysis and post-selection inference. Beyond this, he also gathered experience on applied problems, such as Bayesian modelling of CRISPR data and recurrent neural networks.