Local Influence Analysis for Latent Variable Models with Non-Ignorable Missing Responses
Chapter in an edited book (author)

Times Cited
Altmetrics Information

Other information
AbstractThis chapter provides a general procedure for the local influence analysis of a generic latent variable model with non-ignorable missing data. The local infl{ligature}uence measures are based on the conditional expectation of the complete-data log-likelihood function in the corresponding EM algorithm. The proposed methodology is feasible for a wide variety of perturbation schemes. Especially, a minor perturbation to the missing model is introduced to investigate the effect of how minor perturbations on the missing mechanism model can lead to a large impact on the key features of the whole model. For illustration of the methodology, two models are considered. For the normal mixed effects model, results that are obtained from analyses of a simulation study, and a real example on a longitudinal study of a kidney disease are presented. While for the generalized linear mixed model, an artificial example is used to show the performance of the method. © 2007 Elsevier B.V. All rights reserved.
All Author(s) ListBin L., Song X.-Y., Lee S.-Y., Lai F.M.-M.
All Editor(s) Listed. by Sik-Yum Lee.
Detailed descriptioned. by Sik-Yum Lee.
Pages109 - 134
LanguagesEnglish-United Kingdom

Last updated on 2020-02-07 at 02:47