| ||||
| ||||
![]() Title:A Bootstrap Adaptive LASSO Estimator for Longitudinal Data: a Stability-Oriented Monte Carlo Study Conference:TMREES26Lb Tags:Adaptive LASSO, bootstrap, longitudinal data, regularization, repeated measures and variable selection Abstract: Longitudinal studies generate repeated observations on the same subjects and therefore combine variable-selection difficulty with within-subject dependence. Standard LASSO can select sparse models, but its selection may be unstable in finite samples, while Adaptive LASSO improves coefficient-specific penalization without directly quantifying selection stability. This paper develops a bootstrap-stabilized Adaptive LASSO estimator for longitudinal linear models. Subjects, rather than individual rows, are resampled to preserve the repeated-measures structure. Adaptive LASSO is fitted within each bootstrap sample, selection frequencies are computed, and variables exceeding a stability threshold are retained; final coefficients are aggregated by the bootstrap median. A Monte Carlo experiment considers 12 candidate predictors, four truly active effects, three sample sizes (50, 100, 200 subjects), five repeated measurements, and AR(1) within-subject correlations of 0.2, 0.6, and 0.8. The proposed procedure preserved all active predictors in the investigated scenarios and substantially reduced false-positive selections. A Bootstrap Adaptive LASSO Estimator for Longitudinal Data: a Stability-Oriented Monte Carlo Study ![]() A Bootstrap Adaptive LASSO Estimator for Longitudinal Data: a Stability-Oriented Monte Carlo Study | ||||
| Copyright © 2002 – 2026 EasyChair |
