Evaluation of bias in systematic reviews (R: Review)

• Antragsteller: Dr. Guido Schwarzer
• Fachliche Zuordnung: Epidemiologie und Medizinische Biometrie/Statistik
• Förderung: Förderung von 2004 bis 2011
• Projektkennung: Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 5470786

The validity of meta-analyses may be affected by various sources of bias, including publication bias as a predominating factor. The development of statistical methods for the detection and modeling of publication bias in meta-analysis has been the topic of much methodological research over recent years. Most prominent methods to model publication bias are selectivity models and selection models, which are based on weighted distribution theory. Due to the complexity of these methods only few applications have been published, most by the developers themselves. A comprehensive comparison of these methods both in applications and simulations has not been conducted, so far. The properties of selectivity models will be investigated in this project both in ongoing meta-analyses conducted by members of this group and in simulations. Furthermore, the results of meta-analyses of non-randomized studies based on individual patient data and meta-analyses of non-randomized studies based on aggregated data will be compared in practical examples.

• DFG-Verfahren: Forschungsgruppen
• Teilprojekt zu: FOR 534: Statistical Modeling and Data Analysis in Clinical Epidemiology