When is the meeting? | Wednesday 10-May-2023 at 10.00 - 11.00 (NB! unusual week day!) |
| Where is the meeting? | Zoom video-conference: http://meet.nbis.se/aiaio |
| What is the meeting about? |
Uncertainty quantification for clinical decision-making | |
| Who is the responsible presenter? | Dave |
Zachariah, Div. Systems & Control, Dept. Information Technology, Uppsala University |
Abstract: TBA In clinical decision-making problems, one has to assess the impact of different treatment options on patients with varying characteristics. Any decision policy determines the choice of treatments based on such characteristics. A fundamental problem in machine learning and causal inference is to evaluate performance of alternative decision policies using data.
In this talk we will consider the problem of making statistically valid inferences about the out-of-sample performance of a given decision policy when the training data is observational rather than experimental. We show that it is possible to draw such inferences with finite-sample coverage guarantees, and propose a method to certify the performance of decision policies policy using observational data under any specified range of credible model assumptions.
Presentation: TBA (after session)