"Marginal Analysis of Longitudinal Data in Presence of Non-Ignorable Missing Data"
Haiqun Lin, Ph.D., Yale University
April 20, 2010 @ 3:30 pm - 4:30 pm
Location: BRB 253
Biostatistics
Title: Marginal Analysis of Longitudinal Data in Presence of Non-Ignorable
Missing Data.
Abstract: We proposed a method of latent variable prediction for population inference of longitudinal data in the presence of non-ignorable missing data. For population inference, generalized estimating equations (GEE) provide a valid approach under the situation of missing completely at random (Liang and Zeger 1986). For the situation of missing at random, the weighted estimating equation proposed by Robins and his colleagues is valid approach (Robins, Rotnitzky and Zhao 1995). For non-ignorable missing data, we integrate the latent variable modeling approach as an individualized prediction of missing data probability into weighted GEE. Our estimator of population parameter is consistent. Our method is evaluated with simulation and illustrated with a longitudinal anti-psychotic trial with a large fraction of dropouts.
