The statement verdict table. Distinguishing-for-a-factor is the joint
event that every contrast touching that factor exceeds the posterior
critical difference delta_kl (computed from the posterior spread of
the score contrasts); consensus is the equivalence event that all
scores sit within one grid column (delta_grid()) of each other. Both
families are selected through the posterior false-discovery rule, and a
statement can be distinguishing, consensus, or indeterminate.
Arguments
- fit
A
bayesqm_fitwithK >= 2.- q
Posterior expected false-discovery bound (default 0.05).
- cons_floor
Minimum consensus probability before a statement can be selected as consensus (default 0.95).
- prob
Credible-interval probability for the contrast intervals; defaults to the fit's. The critical difference itself keeps its z_0.975 definition regardless.
Value
A data frame with one row per statement: per-factor
distinguishing probabilities, consensus_prob, and the verdict.
Attributes: delta_kl and delta_kl99 (per pair), delta_grid,
contrasts (the per-pair contrast table with intervals, the
probability diff_column_prob that the two viewpoints place the
statement in different grid columns, and stars, * for selected
claims and ** for those that also clear the z_0.995 critical
difference at probability .99), and expected_false per family.
Examples
compute_qdc(demo_fit())
#> statement f1_dist_prob f2_dist_prob consensus_prob verdict
#> 1 S1 0.300 0.300 0.250 indeterminate
#> 2 S2 0.160 0.160 0.450 indeterminate
#> 3 S3 0.170 0.170 0.445 indeterminate
#> 4 S4 0.940 0.940 0.005 distinguishing (f1, f2)
#> 5 S5 0.060 0.060 0.645 indeterminate
#> 6 S6 0.915 0.915 0.005 distinguishing (f1)
#> 7 S7 0.100 0.100 0.500 indeterminate
#> 8 S8 0.195 0.195 0.315 indeterminate
#> 9 S9 0.135 0.135 0.430 indeterminate
#> 10 S10 0.540 0.540 0.125 indeterminate
#> 11 S11 0.980 0.980 0.000 distinguishing (f1, f2)
#> 12 S12 0.040 0.040 0.630 indeterminate
#> 13 S13 0.370 0.370 0.180 indeterminate