Refit omitting each study in turn. Refitting is the point: deleting
a study changes tau^2 and therefore ALL the weights, so a deletion
can move the pooled estimate far more than that study's own weight
suggests – the channel a quick recomputation with the full-data
tau^2 misses. The diagnostic worth reading is whether any single
deletion changes a CONCLUSION, which flips_significance records.
Usage
morie_psy_ma_loo(yi, vi, method = "PM")
Arguments
- yi, vi
effects and within-study variances.
- method
heterogeneity estimator for each refit: "PM",
"REML" or "DL".
Value
list: mu_full, tau2_full, mu_loo, tau2_loo, delta_mu,
ci_loo, flips_significance, most_influential, significant_full,
method_used, k, method.
References
Viechtbauer and Cheung (2010), Research Synthesis
Methods 1:112-125.
Examples
morie_psy_ma_loo(c(0.2, 0.5, 0.1, 0.4), c(0.02, 0.03, 0.04, 0.02))$mu_full
#> [1] 0.3100343
