Browse Teach Curate

Penalized Quasi-likelihood via Within-Cluster Resampling for Informative Cluster Size

A high-dimensional variable-selection method for longitudinal data when cluster size is informative: draw one observation per cluster, fit a SCAD-penalized quasi-likelihood model to each resampled independent dataset, and aggregate the estimators with a penalized mean-regression step, achieving model-selection consistency where GEE is biased.

Dr. Theopolis Text

Penalized Quasi-likelihood for High-dimensional Longitudinal Data via Within-cluster Resampling Yue

Proposes a penalized quasi-likelihood method for simultaneous variable selection and estimation in high-dimensional longitudinal (clustered) data when the cluster size is informative (ICS) - i.e. the…