الفهرس 
Title : Sensitivity Analysis in Longitudinal Data with Missing Values
AbstractOnly 14 pages are availabe for public view
 |  | from | 114 |  |  |
| | | from | 114 | | |
Abstract
In longitudinal studies outcome variable(s) are measured for the same individual on several different occasions. When the intended measurement for an individual could not be obtained, missing data arise and this creates a major challenge for data analysis. Models, such as selection models, pattern mixture models and shared parameter models, are used in longitudinal analysis to handle non-ignorable missingness. Shared parameter model is intuitively appealing for biomedical researches because there are always some unobserved variables responsible for a person{u2019}s vulnerability to death or disease. Though, these models are based on assumptions about the missing data mechanism that cannot be verified by the observed data.An appropriate approach to the problem of the unverifiable assumptions is to conduct a sensitivity analysis. One of the tools of the local sensitivity is the Index of Sensitivity to Non-Ignorability (ISNI) proposed by Troxel et al. (2004). This index measures the local sensitivity of the parameter estimate to departures from ignorability. Most of the studies in the previous literature derived the ISNI for selection models. However, Viviani (2012) extended the ISNI in the context of joint modeling of the longitudinal and the survival data. This study proposes a modified Stochastic-EM algorithm to obtain the parameters{u2019} estimates of the shared parameter model. Furthermore, a modified procedure to find the standard error the parameters{u2019} estimates is developed. A simulation study is performed to evaluate the performance of the proposed approach. Regarding the sensitivity of parameters{u2019} estimates of the measurement model to the assumption of non-ignorable dropout, the study proposes an extension to ISNI in the context of joint modeling of the longitudinal and the missing data.The formula of the extended ISNI is derived and a simulation study is conducted to investigate the sensitivity of the model parameter estimates to missing data assumptions using the ISNI