B1571
Title: Missing data with causal and statistical dependence
Authors: Elizabeth Ogburn - Johns Hopkins University (United States) [presenting]
Abstract: Two recent projects on causal inference in the presence of missing data are described. In one project, spatial dependence is harnessed to help create proxies for missing confounders. In this setting, the presence of statistical dependence is assumed to lend structure to the unmeasured confounder, and this structure facilitates the identification of causal effects. In the other project, a new kind of missing data process is identified, in which missingness indicators can exhibit causal dependence across units; this kind of dependence undermines existing identification results for missing data and requires new graphical model-based results.
