MNAR Imputation

How it works

Benefits

Preparation

Risks & contraindications

Evidence quality

Highlighted claims

• The imputation stage asks the LLM to propose executable Python imputation operators rather than directly fill missing cells. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• In the formal setting, covariate missingness is MNAR because missingness may depend on unobserved values, other covariates, treatment, and outcome. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• The evolutionary imputer mutates a deterministic seed imputer for 20 iterations. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• Candidate imputers are scored using RMSE plus penalties for biomarker-outcome and biomarker-treatment correlation distortion. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• Runtime and static guards reject unsafe or invalid imputation candidates. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• LLM-driven imputation methods had the strongest pooled performance under MNAR imputation benchmarks. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• Electronic health record analyses face high-rate missing-not-at-random biomarker missingness alongside time-varying confounding. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation
• The article argues that imputation in causal pipelines should be evaluated using both biomarker metrics and causal metrics. — Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation