Tossing coins: design of sequential experiments with covariates
This seminar has been postponed due to unforeseen circumstances. We apologise for any inconvenience this may cause.
Randomized experiments are considered the gold standard of study designs. When averaged over all possible randomizations, the treatment effect is free from bias and can also have a causal interpretation. In many experiments in the social and medical sciences, participants become available sequentially over time. Covariates are taken when the patient arrives and the treatment is assigned soon after. In this sequential setting, allocating treatments completely at random can lead to imbalance in the replication of treatments, as well as treatments being unbalanced for covariates. Both problems can lead to imprecision in the estimate of the treatment effect.
Sequential optimal design based methodology allocates treatments to minimize the variance of the treatment effect under a specified model. We extend the optimal design based methodology to a nonmyopic setting, where treatment allocation for the current patient depends not only on the treatments and covariates of the patients in the study, but also the impact of possible future treatment assignments. One main limitation of the nonmyopic approach is that it involves computationally expensive recursive formulae which can only be implemented in limited contexts. This motivated the development of a pseudo-nonmyopic approach which has a similar aim to the nonmyopic approach, but does not involve recursion. We illustrate through simulation studies in the setting of personalized medicine that the myopic approach is often a more efficient approach than the nonmyopic or pseudo-nonmyopic approaches.
This is joint work with Dave Woods, Kim May Lee, Ilya Shpitser and Peter Smith.
About the speaker
Mia Tackney is a Research Fellow in the Medical Statistics department working with Elizabeth Williamson and James Carpenter on missing data methods for clinical trials that incorporate data from wearable devices. Previously, she did her Ph.D. in the design of experiments group at the University of Southampton which investigated nonmyopic approaches to sequential design in the presence of covariates. She was supervised by Dave Woods, Ilya Shpitser and Peter Smith.
Please note that this session will NOT be live-streamed/recorded