Skip to contents

Compute the realized mean informativeness \(\widehat{I} = \sigma_\tau^2 / (\sigma_\tau^2 + \mathrm{GM}(\widehat{se}_j^2))\), where GM is the geometric mean of the per-site sampling variances. Higher \(\widehat{I}\) indicates more precise per-site estimates relative to the between-site heterogeneity scale; a feasibility index near 1 means each site's estimate alone identifies the latent effect.

Usage

compute_I(se2_j, sigma_tau, tau_j = NULL)

Arguments

se2_j

Numeric vector (length \(\ge 2\)). Per-site sampling variances \(\widehat{se}_j^2\). Must be strictly positive.

sigma_tau

Numeric (\(\ge 0\)). Between-site SD. 0 returns 0 (no heterogeneity → no informativeness to recover).

tau_j

Optional numeric vector matching se2_j length. Kept for signature compatibility with the S3 method; not used in the computation.

Value

A scalar double in [0, 1).

Details

This is the realized counterpart to the design-target \(I\) fed to sim_meta or used in gen_se_direct. Under the deterministic-grid direct-precision path, the realized and target \(I\) match exactly; under the site-size-driven path or under a custom se_fn, the realized value is what you read off the simulation.

tau_j is accepted for data-method signature compatibility (so the same call shape works for multisitedgp_data objects and bare numeric vectors) but does not enter the statistic.

References

Lee, J., Che, J., Rabe-Hesketh, S., Feller, A., & Miratrix, L. (2025). Improving the estimation of site-specific effects and their distribution in multisite trials. Journal of Educational and Behavioral Statistics, 50(5), 731–764. doi:10.3102/10769986241254286 .

See also

informativeness for the S3 generic with methods for multisitedgp_data and numeric se2_j inputs; compute_kappa for the precision constant input to the site-size-driven path; heterogeneity_ratio for the realized \(R\).

Other family-diagnostics: bhattacharyya_coef(), compute_kappa(), compute_shrinkage(), default_thresholds(), feasibility_index(), heterogeneity_ratio(), informativeness(), ks_distance(), mean_shrinkage(), realized_rank_corr(), realized_rank_corr_marginal(), scenario_audit()

Examples

se2 <- c(0.05, 0.08, 0.10)
compute_I(se2, sigma_tau = 0.20)
#> [1] 0.3518629