= Leave-one-study-out influence analysis
{title2=$\widehat\mu-\widehat\mu_{(-i)}$}
Refit a <meta-analysis> after deleting each study, comparing fitted effects, uncertainty and heterogeneity to the full fit. For fixed weights $w_i>0$, total weight $W$ and weighted effect $\widehat\mu$, direct subtraction of the two weighted means gives $\widehat\mu-\widehat\mu_{(-i)}=w_i(y_i-\widehat\mu)/(W-w_i)$. Re-estimating heterogeneity additionally measures influence through the weights. This is a sensitivity diagnostic, not an automatic rule for removing discordant studies.
Back to article page
