μ^±tk−2α√{τ^2+SE^(μ^)2}, where tk−2α is the 100(1−α/2)% percentile of the t-distribution with k−2 degrees of freedom.[57] It becomes implicit that when an association has been probed in only a few datasets, then the prediction interval will be wider than the respective confidence interval, even if there is no demonstrable between-study variance (i.e. τ2=0). Table 1 summarizes some issues that arise in selecting, interpreting and comparing the properties and results of various commonly used meta-analyses methods.
