This is an archived version of the Handbook. For the current version, please go to training.cochrane.org/handbook/current or search for this chapter here.

Other authors have proposed more sophisticated methods that avoid strong assumptions about the association between study P value and publication probability (Dear 1992, Hedges 1992). These methods can be extended to estimate intervention effects, corrected for the estimated publication bias (Vevea 1995). However they require a large number of studies so that a sufficient range of study P values is included. A Bayesian approach in which the number and outcomes of unobserved studies are simulated has also been proposed as a means of correcting intervention effect estimates for publication bias (Givens 1997). Recent work has examined the possibility of assessing robustness over a range of weight functions, thus avoiding the need for large numbers of studies (Vevea 2005). The complexity of the statistical methods, and the large number of studies needed, probably explain why selection models have not been widely used in practice.