Identifiability of total effects from abstractions of time series causal graphs
This work addresses a practical challenge in causal inference for time series data, offering theoretical guarantees for researchers and practitioners dealing with incomplete graph information.
The paper tackles the problem of identifying total effects from interventions in observational time series when only abstracted causal graphs are available, showing that identifiability is always possible for extended summary graphs and providing conditions and adjustment sets for summary graphs.
We study the problem of identifiability of the total effect of an intervention from observational time series in the situation, common in practice, where one only has access to abstractions of the true causal graph. We consider here two abstractions: the extended summary causal graph, which conflates all lagged causal relations but distinguishes between lagged and instantaneous relations, and the summary causal graph which does not give any indication about the lag between causal relations. We show that the total effect is always identifiable in extended summary causal graphs and provide sufficient conditions for identifiability in summary causal graphs. We furthermore provide adjustment sets allowing to estimate the total effect whenever it is identifiable.