ABSTRACT
Publication bias occurs when some studies are systematically missing in the published literature, and thus not included in a meta-analysis. Publication bias, along with other reporting biases and questionable research practices (QRPs), can distort the results of meta-analyses.
This chapter introduces a general taxonomy of reporting biases, and covers several statistical techniques that address publication bias. First, small-study effect methods are described, including funnel plots, Egger's and Peter's test, Duval & Tweedie's trim-and-fill method, PET-PEESE, and limit meta-analyses. Other covered topics are the P-curve method, and (three-parameter) selection models.
Lastly, the overall advantages and disadvantages of different publication bias methods are discussed. All covered methods are illustrated using hands-on examples in R.
