Forest plots from meta-analyses across 2016 CRA, 2017 CRA and 2017 CRG datasets.

Sub-panels show the effects of the following on logged trial latency: a) Plenty Amount; b) Hard Effort; c) ratio requirement; d) interaction between ratio requirement and Hard Effort. Shown are the central estimates of effect size and their 95% confidence intervals. FE Model stands for ‘Fixed Effects’ model.