Is the small positive effect reported in the 2015 meta-analysis of ninety retroactive-influence experiments explicable entirely by publication bias and analytic flexibility?

Bem, Tressoldi, Rabeyron and Duggan report a small overall effect across ninety experiments and argue that bias-correction methods do not remove it. Critics answer that the correction methods available cannot distinguish a genuine small effect from a literature filtered by what got written up, particularly where many of the constituent studies were not pre-registered. The dispute is about the adequacy of the statistical tools, and it is live in the wider methodology literature and not only in parapsychology.