Return the canonical quality-gate thresholds consumed by
scenario_audit when scoring a design grid against
Dr. Chen's four-question diagnostic rubric. Each gate maps a Group
A/B/C/D diagnostic to a PASS / WARN / FAIL band, so the audit
output is a single calibrated verdict rather than a raw diagnostic
table.
Value
A named list of five scalar threshold gates:
mean_shrinkage_minNumeric
0.30. Minimum \(\bar{S}\) for Group D PASS — designs below this floor do too little partial-pooling work to benefit from a hierarchical estimator.feasibility_minNumeric
5.0. Minimum Efron feasibility \(\sum_j S_j\) for Group A PASS — corresponds to the equivalent of about five perfectly-informative sites.R_maxNumeric
30.0. Maximum realized heterogeneity ratio \(R\) for Group A PASS — a ceiling on the ill-conditioning of the precision-to-heterogeneity scale.bhattacharyya_minNumeric
0.85. Minimum Bhattacharyya coefficient between target and realized \(G\) for Group C PASS — high-overlap shape recovery.ks_maxNumeric
0.10. Maximum Kolmogorov-Smirnov distance between target and realized \(G\) for Group C PASS — tail-sensitive complement to the Bhattacharyya gate.
Details
Reading guide. The five named entries cover three diagnostic groups:
Group A — precision and feasibility.
feasibility_min = 5.0for the Efronfeasibility_index;R_max = 30.0for the realizedheterogeneity_ratio.Group C — distribution recovery.
bhattacharyya_min = 0.85forbhattacharyya_coef;ks_max = 0.10forks_distance.Group D — downstream shrinkage.
mean_shrinkage_min = 0.30formean_shrinkage.
Group B diagnostics (realized rank correlation) are tracked
qualitatively in scenario_audit rather than via a
fixed numerical threshold and so do not appear here. Override the
defaults by editing the returned list and passing it to
scenario_audit(); the returned audit object always records which
thresholds were used so the verdict is reproducible.
For the calibration rationale behind each numeric value see the A3 · Diagnostics in practice vignette.
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
scenario_audit for the audit consumer;
mean_shrinkage for the Group D diagnostic;
feasibility_index for the Group A feasibility
diagnostic;
heterogeneity_ratio for the Group A \(R\)
diagnostic;
bhattacharyya_coef and ks_distance
for the Group C distribution-recovery diagnostics;
the A3 ·
Diagnostics in practice vignette.
Other family-diagnostics:
bhattacharyya_coef(),
compute_I(),
compute_kappa(),
compute_shrinkage(),
feasibility_index(),
heterogeneity_ratio(),
informativeness(),
ks_distance(),
mean_shrinkage(),
realized_rank_corr(),
realized_rank_corr_marginal(),
scenario_audit()
Examples
# Inspect the canonical gates.
default_thresholds()
#> $mean_shrinkage_min
#> [1] 0.3
#>
#> $feasibility_min
#> [1] 5
#>
#> $R_max
#> [1] 30
#>
#> $bhattacharyya_min
#> [1] 0.85
#>
#> $ks_max
#> [1] 0.1
#>
# Tighten the Group D gate before passing to the audit.
th <- default_thresholds()
th$mean_shrinkage_min <- 0.40
th
#> $mean_shrinkage_min
#> [1] 0.4
#>
#> $feasibility_min
#> [1] 5
#>
#> $R_max
#> [1] 30
#>
#> $bhattacharyya_min
#> [1] 0.85
#>
#> $ks_max
#> [1] 0.1
#>