New models improve causal inference for longitudinal data

Researchers have developed two new models, CSSD and CSSPD, to improve causal inference from longitudinal observational data, a crucial task for clinical decision support. These models address a fundamental tension in existing methods where adversarial training for treatment invariance suppresses vital outcome prediction signals. CSSPD, in particular, incorporates contrastive predictive coding and local information maximization to enhance temporal predictability and recover lost covariate information. Experiments on the MIMIC-III database showed CSSPD outperforming the Causal Transformer in counterfactual root mean squared error, and on a Cancer Simulation dataset, CSSD achieved the lowest overall average RMSE. AI

IMPACT These models offer a more robust approach to understanding treatment effects over time, potentially improving clinical decision-making and medical research.

RANK_REASON The cluster contains an academic paper detailing new models and experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• Cancer Simulation
• Causal State-Space model with Direct decoder
• Causal State-Space model with Predictive regularisation and Direct decoder
• Causal Transformer
• CSSPD
• MIMIC-III

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →