Resumen de Improving resampling techniques for linear mixed models: nested and adjusted semiparametric bootstrap

José Antonio Sánchez Espigares, Jordi Ocaña i Rebull

• Resampling methods for mixed-effects models extend the capabilities of inference when modeling clustered data, such repeated measures and longitudinal data. The classical approach to the semiparametric bootstrap works with the empirical distribution of the random-effects and residuals. We suggest a modified semiparametric bootstrap as a result of combining two features: adjusting the sample to the estimated parameters and keeping the nested structure for the different levels when bootstrapping. A simulation study has been implemented in S-Plus in order to compare the different bootstrap methodologies.