points
However, the estimates are still probably overestimated. Confounding, p-hacking, publication bias, all move us towards larger estimates.