Longitudinal Data with Intermittent Missing Values: A Sensitivity Analysis Approach | ||||
| The Egyptian Statistical Journal | ||||
| Article 1, Volume 51, Issue 2, December 2007, Page 37-47 PDF (14.72 MB) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/esju.2007.313448 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Ahmed M. Gad  1; Abeer S. Ahmed* 2
| ||||
| 1Statistics Department, Faculty of Economics & Political Science, Cairo University | ||||
| 2The National Center for Social and Criminology Research, Cairo, Egypt | ||||
| Abstract | ||||
| Intermittent missing data are not uncommon in longitudinal data studies. In selection models, the probability of being missing for any observation is modeled as a function of the current observation and the previous observations. The parameter that relates the probability of missingness and the current observation has special interpretation. The degree of informativeness of the missing data process depends on this parameter's value. We conduct sensitivity analysis to evaluate the effect of this parameter value (the sensitivity parameter) on study results. In the proposed approach, the sensitivity parameter is assumed to be fixed at a set of plausible values. This allows us to examine several degrees of informativeness of the missing data process. The stochastic EM algorithm is used to obtain parameter estimates. The proposed method is evaluated via a simulation study and then applied to a real data set. Sensitivity analysis shows that the conclusion depends on the degree of informativeness. Hence, when estimating the sensitivity parameter the results should be interpreted cautiously. | ||||
| Keywords | ||||
| Diggle; Kenward Model; Informative Missing; Intermittent Missing; Selection Models; The Stochastic EM Algorithm | ||||
|
Statistics Article View: 93 PDF Download: 10 |
||||
