Ridderprint

Fostering overseas success: A meta-analysis 105 5.5.2.4 Publication Bias Using the “trim and fill ” method (Duval & Tweedie, 2000), we estimated, per criterion, the number of missing studies due to publication bias. Typically, studies with small samples or nonsignificant results do not meet the requirements for publication. This causes “ research that appears in the published literature [to be] systematically unrepresentative of the population of completed studies ” (Rothstein, Sutton, & Borenstein, 2006, p. 1). We estimated the average true effect sizes before and after correcting for potential bias. Table 5.10 and Figure 5.2 suggested minor publication bias. For retention, an estimated five effect sizes were missing which would have decreased the true effect slightly, from ρ = .21 to ρ = .17. For performance outcomes, six studies with larger effect sizes were estimated missing and correcting for these increased the estimated true effect from ρ = .19 to ρ = .22. Figure 5.2. Funnel plots of the estimated extent of publication bias in literature. Grey dots represent the observed data points; black dots are estimated missing effect sizes.