Title: Robust inference in a heteroskedastic multilevel linear model with structural change
Authors: Ronrick Da-ano - University of the Philippines-Diliman (Philippines) [presenting]
Erniel Barrios - University of the Philippines (Philippines)
Joseph Ryan Lansangan - University of the Philippines (Philippines)
Abstract: A heteroskedastic multilevel model with cross-sectional interactions in higher and lower levels with structural change is estimated by a hybrid procedure of the forward search algorithm preceding bootstrap method. The simulation study exhibits the ability of the hybrid procedure to produce narrower confidence interval even when there is model misspecification error and structural change. Moreover, it has a comparable predictive ability with the classical restricted maximum likelihood (REML) estimation. However, the hybrid method yield estimates of the parameters with lower bias relative to REML. The hybrid of forward search and bootstrap method can further robustify estimates of fixed and random coefficients under various levels of interclass correlation or in the presence of structural change in a multilevel model.