Statistics Colloquium : Dr. Guido Knapp

Abstract: Publication bias is a serious problem  in meta-analysis of aggregated data. If not all studies for the combination of results are available, the overall result of the meta-analysis will be biased. Approaches of dealing with publication bias can be distinguished into two classes: selection models and funnel-plot-based methods. We will discuss these methods in this talk and illustrate them by various real-data examples. Some comparisons of the methods via simulation results will be indicated.