Biostatistics Journal Club: Multiple Imputation by Super Learning (MISL)

In the presence of missing data, multiple imputation methods such as Multiple Imputation by Chained Equations (MICE) are widely used but depend on correct specification of imputation models. This talk presents Multiple Imputation by Super Learning (MISL), an ensemble-based extension that flexibly combines parametric and nonparametric learners to better handle missingness within complex data structures. This talk will compare MISL to standard multiple imputation approaches and show that MISL can reduce bias and improve confidence interval coverage, often with comparable or narrower interval widths.

Required Reading:

MISL: Multiple Imputation by Super Learning

Speaker:

Justin Manjourides, PhD, Northeastern University