Statistics Colloquium: Xiaotong Shen (University of Minnesota)

Date: 

Monday, April 5, 2021, 10:30am to 11:30am

Location: 

Zoom - please contact emilie_campanelli@fas.harvard.edu for more information

Title:

Inference for a directed acyclic graphical model with interventions

Abstract:

Consider an inference problem in a directed acyclic graphical model subject to unknown interventions. In this presentation, we will give conditions for multiple unknown interventions to yield an identifiable model. For inference, we identify the ancestral relations and the interventions for each hypothesis-specific primary variable.
Towards this end, we propose a likelihood ratio test based on data perturbation to account for the identification effect by perturbing original data to assess the uncertainty associated with identifying ancestors and interventions. For testing the presence and strengths of parent-child relations in a pathway,  we show that the proposed tests achieve desired statistical properties in terms of controlling Type I and Type II errors for a large graph.  Numerical examples will be given to demonstrate the utility and effectiveness of the proposed procedure.
This work is joint with Chunlin Li and Wei Pan of the University of Minnesota.