docs/qa_reports/miami/__cross_publisher_race.json
Served verbatim from the project repository. Internal working document conventions apply: documents may reference file paths, branch names, and findings-ledger anchors from the repo.
{
"summary": {
"years": [
"2018-19",
"2022-23",
"2023-24"
],
"verdict": "PASS",
"uniform_split_school_years": [],
"verdict_note": "Served FLDOE-derived share vs NCES CCD race subtotal. (a) CONSISTENCY, not accuracy: |\u0394| within the process bound min(9 \u00d7 hidden cells, unallocated) / total + 0.5 pp for \u2265 99% of school-years, per share, with hidden cells counted as the loader counts them \u2014 any allocation that respects the cap and sums to the total satisfies it (the plain equal split included), so a failure means a loader defect, not a poor prior; (b) no school-year served with all seven shares equal; (c) ACCURACY against the no-information reference: the served share's mean |\u0394| from CCD is no larger than the equal split's (the two coincide except where the capped adjacent-year prior applies, 2022-23). The prior was chosen among five candidates on CCD 2022-23, the same reference reported here (in-sample selection; the held-out split-half check is in docs/cities/miami/probe/race_imputation_prior.py). Mean |\u0394| and the within-2-pp share are informational."
},
"by_metric": {
"pct_black": {
"n_pairs": 1533,
"within_bound": 1533,
"within_bound_share": 1.0,
"within_2pp": 1385,
"within_2pp_share": 0.903457273320287,
"median_delta": 0.0,
"mean_abs_delta": 0.8758065558266108,
"mean_abs_delta_equal_split_baseline": 1.38016555829809,
"beats_equal_split": true,
"verdict": "PASS",
"largest": [
[
"7861",
"2022-23",
19.64,
62.5,
100.5
],
[
"4651",
"2018-19",
33.33,
72.73,
82.32
],
[
"7835",
"2018-19",
43.75,
67.31,
100.5
],
[
"7892",
"2023-24",
50.0,
72.22,
100.5
],
[
"7840",
"2018-19",
66.67,
87.18,
100.5
],
[
"7811",
"2022-23",
39.39,
57.89,
100.5
],
[
"7805",
"2023-24",
75.0,
92.86,
100.5
],
[
"8012",
"2023-24",
45.45,
27.78,
100.5
]
]
},
"pct_hispanic": {
"n_pairs": 1533,
"within_bound": 1533,
"within_bound_share": 1.0,
"within_2pp": 1375,
"within_2pp_share": 0.8969341161121983,
"median_delta": 0.0,
"mean_abs_delta": 0.7679654293623916,
"mean_abs_delta_equal_split_baseline": 1.2701901124913315,
"beats_equal_split": true,
"verdict": "PASS",
"largest": [
[
"7861",
"2022-23",
73.81,
37.5,
100.5
],
[
"7855",
"2022-23",
50.0,
81.82,
100.5
],
[
"6997",
"2018-19",
55.56,
82.61,
100.5
],
[
"5417",
"2023-24",
33.33,
56.25,
100.5
],
[
"8012",
"2023-24",
45.45,
66.67,
100.5
],
[
"6997",
"2023-24",
56.47,
77.27,
75.95
],
[
"7840",
"2018-19",
33.33,
12.82,
69.73
],
[
"7811",
"2022-23",
30.68,
10.53,
100.5
]
]
},
"pct_white": {
"n_pairs": 1533,
"within_bound": 1533,
"within_bound_share": 1.0,
"within_2pp": 1350,
"within_2pp_share": 0.8806262230919765,
"median_delta": 0.0,
"mean_abs_delta": 0.9327424846746709,
"mean_abs_delta_equal_split_baseline": 1.2347621314110193,
"beats_equal_split": true,
"verdict": "PASS",
"largest": [
[
"7855",
"2022-23",
25.0,
9.09,
100.5
],
[
"7892",
"2023-24",
25.0,
11.11,
100.5
],
[
"8181",
"2023-24",
23.43,
9.9,
62.88
],
[
"6997",
"2023-24",
29.02,
17.27,
75.95
],
[
"7863",
"2018-19",
14.29,
4.55,
41.41
],
[
"7855",
"2023-24",
25.0,
15.38,
100.5
],
[
"1019",
"2022-23",
9.52,
18.57,
23.36
],
[
"8121",
"2018-19",
12.65,
3.7,
33.83
]
]
}
},
"by_year": {
"2018-19": {
"hidden_cells": 5700,
"tokens_resolved_as_zero": 0,
"by_metric": {
"pct_black": {
"n_pairs": 500,
"within_bound_share": 1.0,
"within_2pp_share": 0.904,
"mean_abs_delta": 0.9053416488150688,
"mean_abs_delta_equal_split_baseline": 0.9053416488150688,
"lift_vs_equal_split": 0.0
},
"pct_hispanic": {
"n_pairs": 500,
"within_bound_share": 1.0,
"within_2pp_share": 0.884,
"mean_abs_delta": 0.8236294145444824,
"mean_abs_delta_equal_split_baseline": 0.8236294145444824,
"lift_vs_equal_split": 0.0
},
"pct_white": {
"n_pairs": 500,
"within_bound_share": 1.0,
"within_2pp_share": 0.872,
"mean_abs_delta": 0.9358186863691705,
"mean_abs_delta_equal_split_baseline": 0.9358186863691705,
"lift_vs_equal_split": 0.0
}
}
},
"2022-23": {
"hidden_cells": 15982,
"tokens_resolved_as_zero": 1482,
"by_metric": {
"pct_black": {
"n_pairs": 515,
"within_bound_share": 1.0,
"within_2pp_share": 0.9106796116504854,
"mean_abs_delta": 0.8902606897707437,
"mean_abs_delta_equal_split_baseline": 2.3915856427586615,
"lift_vs_equal_split": 0.6277529544190439
},
"pct_hispanic": {
"n_pairs": 515,
"within_bound_share": 1.0,
"within_2pp_share": 0.9223300970873787,
"mean_abs_delta": 0.6551580732219718,
"mean_abs_delta_equal_split_baseline": 2.150129799895107,
"lift_vs_equal_split": 0.6952937105220656
},
"pct_white": {
"n_pairs": 515,
"within_bound_share": 1.0,
"within_2pp_share": 0.9029126213592233,
"mean_abs_delta": 0.8358901813087868,
"mean_abs_delta_equal_split_baseline": 1.734911770525917,
"lift_vs_equal_split": 0.5181944145462815
}
}
},
"2023-24": {
"hidden_cells": 5961,
"tokens_resolved_as_zero": 0,
"by_metric": {
"pct_black": {
"n_pairs": 518,
"within_bound_share": 1.0,
"within_2pp_share": 0.8957528957528957,
"mean_abs_delta": 0.8329273560670405,
"mean_abs_delta_equal_split_baseline": 0.8329273560670405,
"lift_vs_equal_split": 0.0
},
"pct_hispanic": {
"n_pairs": 518,
"within_bound_share": 1.0,
"within_2pp_share": 0.8841698841698842,
"mean_abs_delta": 0.8263897456196717,
"mean_abs_delta_equal_split_baseline": 0.8263897456196716,
"lift_vs_equal_split": -2.220446049250313e-16
},
"pct_white": {
"n_pairs": 518,
"within_bound_share": 1.0,
"within_2pp_share": 0.8667953667953668,
"mean_abs_delta": 1.0260645607097685,
"mean_abs_delta_equal_split_baseline": 1.0260645607097685,
"lift_vs_equal_split": 0.0
}
}
}
}
}