Sujit K. Ghosh
North Carolina State University and SAMSI
Analysis of irregularly observed longitudinal data within a mixed model
framework often are based on stringent parametric or semiparametric
model assumptions that are often too restrictive in practice. In
particular, finite dimensional (parametric) models are found inadequate
to address the complex relationship between the response and predictors.
A majority of the currently available models and associated estimation
methodologies are based on restrictive assumptions on the correlation
structure of longitudinal data. To begin with, a flexible class of
non-parametric models based on a sequence of Bernstein polynomials with
varying degrees is developed and the degree of the polynomials is chosen
using a newly proposed predictive divergence criteria. Various simulated
data scenarios are used to illustrate the superior performance of the
proposed estimation and model selection methodology. The newly proposed
models and associated inference are illustrated using real data analysis
of various growth curves.
[This is a joint work with Dr. Liwei Wang, PPD Inc.]
---
All statistics students are expected to attend.