University of Ioannina
Putting the pieces together: accounting for missing data in aggregate and individual patient data meta-analysis
Abstract
dc:descriptionThis dissertation aims to study and develop methodology for meta-analytic models in the presence of missing data, both at the aggregate and individualparticipant levels. Missing data are a common phenomenon in statistical analysis and, consequently, in meta-analysis. Their handling requires the formulation of specific assumptions, which directly influence the modelling process. The introduction presents the fundamental principles of meta-analysis, the theory of missing data, and the available methods for handling missingness within the meta-analytic framework. Subsequently, statistical methodology is developed for the imputation of missing values in individual participant data, through an appropriate modification of the multiple imputation procedure. The central idea underlying this methodology relates to the heterogeneity observed across studies synthesized in a meta-analysis. Given that such heterogeneity may be substantial, it was deemed necessary to develop an approach in which missing values are primarily imputed by borrowing information from studies that are ‘close’ in terms of their characteristics. Finally, with regard to missing data in aggregate-level analyses, both in pairwise meta-analysis and in network meta-analysis, an R package was developed. The package builds upon existing methodologies that employ pattern-mixture models and allows the incorporation of expert knowledge from the relevant field when addressing missing data. The aim of the package is to disseminate the existed methodology and to facilitate its practical implementation by researchers and analysts.
Degree
thesis:*- Grantor dc:publisher
- University of Ioannina
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Christogiannis, Christos
- Χριστογάννης, Χρήστος
Subjects
dc:subject × 10Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/60758