Estimators for a Joint Model of a Longitudinal Process and Informative Time Schedule

• Activity Number: 27
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Lin Yen-Kuang *+ and Michael Bronsert and Khalil Shafie and Siamak Noorbaloochi
• Companies: UNCO and University of Colorado at Denver and Shahid Beheshti University and VA
• Keywords: Longitudinal data ; Informative Time Schedule ; Maximum likelihood estimates

Abstract

In an informative schedule study where the subsequent observations are scheduled base on prior responses, time between observations could be informative. Consequently, traditional fixed time approaches fail to address the informative feature of the data and further attenuating the inferences achieved by these approaches. In this paper we develop a joint model of a longitudinal process and informative time schedule data. Two special cases of the informative schedule models' log likelihood functions and corresponding gradient functions were developed. Maximum likelihood estimates (MLE) for these models were obtained from Monte Carlo simulated data by employing the Multivariate Newton-Raphson optimization routine implemented in a SAS/IML call statement. Parameter estimates were determined for a few select cases of subject and observation length variations under both balanced and unbalanced observation matrices.