Reference Number: DEVCOM-089

Project Description

This research would focus on developing models that can identify causal relationships between variables when working with datasets that combine textual, visual, and other modalities, where some data points are systematically missing. The challenge lies in distinguishing between missing data mechanisms (missing completely at random vs. missing not at random) and ensuring that causal discovery methods remain robust when entire modalities are absent for certain observations. The work would focus on developing identification strategies and estimation methods that can leverage information across modalities while accounting for the potential bias introduced by systematic missingness patterns.

Technical Skills

• Python
• SQL
• Flask