Prediction interval (meta-analysis)

The confidence interval shows how precisely the average of the true effects is estimated. The prediction interval says where the true effect of a new, comparable study is likely to lie: it also accounts for the variation between studies (τ²) and is therefore wider. It shows how reliable the pooled effect is in practice; that is why it is in the main sentence. Statable calculates it as metafor does, with the Knapp-Hartung adjustment (t distribution with k − 1 degrees of freedom). IntHout and colleagues (2016) recommend the t distribution with k − 2 degrees of freedom when there are few studies; this variant is shown in the result under "For specialists".

Example: Pooled r = .15 [.08, .22], prediction interval [−.05, .34]: a new study might also find a relationship close to zero.

Caution: With few studies the interval is very uncertain.