Bayesian Inference for Sequential Treatments Under Latent Sequential Ignorability

Federico Ricciardi from University College London presents a methodological paper on causal inference for longitudinal treatments. The work focuses on assessing the effect of different treatment sequences on a final outcome, comparing the assumptions of sequential ignorability and latent sequential ignorability. The performance of these assumptions is evaluated using theoretical arguments and simulation studies.

Use Cases

• Evaluating causal assumptions for longitudinal treatment studies based on the described theoretical framework.
• Comparing inference results under sequential ignorability versus latent sequential ignorability assumptions.
• Simulating treatment sequence effects on final outcomes using the described latent variable model.
• Investigating bias in causal estimates when unobservable quantities influence treatment assignment.

Strengths

• The work provides a detailed theoretical framework for causal inference with sequential treatments.
• It includes simulation studies to investigate the performance of different causal assumptions.
• The paper is published under an Open Access (green) license, facilitating dissemination.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset appears to be a methodological paper; the availability of actual data files is unclear.

Provenance

Source

University College London

Collection Method

Theoretical and simulation-based research.