Seminar by Alexandra Mossman

Please Note: This seminar will be held in-person.

Overview of Causal Inference in the Presence of Within- and Between-Group Interference

Dynamic treatment regimes (DTRs) are sequences of decision rules that aim to optimize patient treatment at each stage of follow-up. They are frequently applied towards observational datasets under the stable unit treatment value assumption (SUTVA), which states that an individual's outcome is independent of treatments received by others. However, this assumption often does not hold due to different forms of interference that occur within social networks; this is frequently demonstrated in contexts of infectious disease when an individual’s susceptibility to illness can be indirectly affected by whether their neighbours are receiving treatment or vaccination. Much of the biostatistical literature on causal inference considers partial interference, where interference occurs within groups but not across groups, yet much work remains to be explored for modifying DTR methodology that accounts for between-group interference in clustered datasets. This presentation explores a literature review on propensity score-weighted estimators for treatment effects in the presence of interference and proposes a modification of dynamic weighted ordinary least squares regression (dWOLS) to account for individual- and group-specific factors affecting the nature of interference present.