2026 BOY unified subtest and item audit

Math subtests marked primary

Math subtests to monitor / sensitivity

Non-math timed subtests

How to read this page

Timed subtests now include a policy selector: D/trailing-zero is the operational-candidate fixed-form view; Reached/valid-only is a diagnostic/comparability view. STPM is timed and therefore receives both views, but it is labelled Non-math and is not part of the math achievement score.

Excluded from fit means an item was not estimable in that standalone TAM policy view because it had no usable responses or no response variation. It is not a deletion from the raw assessment data.

Contribution labels and rule

Rule	Label
STPM / non-math	Non-math; not in math score
Primary-policy EAP reliability ≥ 0.80 and excluded-from-fit rate ≤ 0.30	Primary contributor
Primary-policy EAP reliability ≥ 0.80 and excluded-from-fit rate > 0.30	Primary; check tail items
0.65 ≤ EAP reliability < 0.80	Contributor; monitor
0.55 ≤ EAP reliability < 0.65	Retain; sensitivity check
EAP reliability < 0.55 or fit incomplete	Review before primary use

Subtest contribution decisions

year_level	test_subgroup	timing	primary_policy	available_policy_views	global_score_contribution_status	decision_basis	eap_reliability	n_items_fit	n_excluded_from_fit	excluded_from_fit_rate	rationale
foundation	STPM	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	non_math_not_in_math_score	Subtest is STPM/non-math. It receives timed policy diagnostics but is excluded from the math achievement score.	0.524	9	0	0.000	Timed non-math subtest; retained for non-math diagnostics, not included in the math achievement score.
foundation	MQ1-20	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	weak_contributor_monitor_sensitivity	0.55 <= primary-policy EAP reliability < 0.65.	0.602	19	11	0.367	Lower standalone reliability; retain for current development fit but require sensitivity checks before promotion.
foundation	MC0-20	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_contributor	Primary policy EAP reliability >= 0.80 and excluded-from-fit item rate <= 0.30.	0.927	50	10	0.167	Standalone reliability strong; retain in math score pending global-model diagnostics.
foundation	MNC0-20	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_contributor	Primary policy EAP reliability >= 0.80 and excluded-from-fit item rate <= 0.30.	0.881	24	6	0.200	Standalone reliability strong; retain in math score pending global-model diagnostics.
foundation	DMT10_2026	Untimed	Valid-only Rasch 1PL	Valid-only	weak_contributor_monitor_sensitivity	0.55 <= primary-policy EAP reliability < 0.65.	0.609	8	3	0.273	Lower standalone reliability; retain for current development fit but require sensitivity checks before promotion.
foundation	BNL0-20	Number Line	NL2 ordinal PCM	NL2 ordinal PCM	contributor_monitor	0.65 <= primary-policy EAP reliability < 0.80.	0.674	10	0	0.000	Moderate reliability; construct-relevant; retain but monitor precision/category/item diagnostics.
year1	STPM	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	non_math_not_in_math_score	Subtest is STPM/non-math. It receives timed policy diagnostics but is excluded from the math achievement score.	0.470	9	0	0.000	Timed non-math subtest; retained for non-math diagnostics, not included in the math achievement score.
year1	MC0-100	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_with_tail_item_sensitivity	Primary policy EAP reliability >= 0.80 but excluded-from-fit item rate > 0.30.	0.940	34	26	0.433	Reliability strong but many tail/constant items under the primary policy; inspect item/global diagnostics.
year1	MNC0-100	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_contributor	Primary policy EAP reliability >= 0.80 and excluded-from-fit item rate <= 0.30.	0.891	22	7	0.241	Standalone reliability strong; retain in math score pending global-model diagnostics.
year1	AAMC	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_contributor	Primary policy EAP reliability >= 0.80 and excluded-from-fit item rate <= 0.30.	0.900	38	2	0.050	Standalone reliability strong; retain in math score pending global-model diagnostics.
year1	ASMC	Timed	D/trailing-zero Rasch 1PL	D/trailing-zero; Reached/valid-only	primary_contributor	Primary policy EAP reliability >= 0.80 and excluded-from-fit item rate <= 0.30.	0.841	25	5	0.167	Standalone reliability strong; retain in math score pending global-model diagnostics.
year1	BNL0-100	Number Line	NL2 ordinal PCM	NL2 ordinal PCM	contributor_monitor	0.65 <= primary-policy EAP reliability < 0.80.	0.727	13	0	0.000	Moderate reliability; construct-relevant; retain but monitor precision/category/item diagnostics.

Difficulty and ceiling check

Short answer: the 2026 BOY math battery does not show a broad operational ceiling problem under the D/trailing-zero fixed-form policy.

Some reached items are easy, especially early magnitude-comparison items.
For timed math subtests, full-form D-frame ceiling rates are zero: unreached trailing items prevent the score from collapsing at the top.
The main local easiness issue is Foundation DMT10_2026: several untimed items are all-correct and the subtest ceiling rate is about 9%.
STPM is timed but non-math, so it is not part of the math achievement score.

Conditional p-valid = accuracy among students who reached/provided a valid response. Operational D p = timed trailing unreached counted as zero. These can tell different stories.

Show technical ceiling/easiness tables

2026 item/subtest ceiling summary

year_level	test_subgroup	n_items	median_conditional_p_valid	items_conditional_p_ge_0.90	median_operational_d_p	items_operational_d_p_ge_0.90	subtest_ceiling_rate_d_sumscore	median_full_form_d_accuracy	median_n_reached_or_valid_count	median_trailing_length	median_nl2_top_category_rate	conclusion
foundation	BNL0-20	10		0		0			10	0	0.251	No NL2 ceiling signal; median exact/near-exact category rate is not high.
foundation	DMT10_2026	11	0.736	4	0.736	4	0.091		11	0		Several easy D-frame items; review item mix.
foundation	MC0-20	60	0.916	33	0.007	5	0	0.233	15	45		Small number of very easy D-frame items; no subtest-level ceiling.
foundation	MNC0-20	30	0.500	0	0.008	0	0	0.167	7	23		Not ceiling-heavy under operational D frame; late/reach effects dominate.
foundation	MQ1-20	30	0.441	2	0.003	0	0	0.1	4	26		Not ceiling-heavy under operational D frame; late/reach effects dominate.
year1	AAMC	40	0.723	9	0.007	0	0	0.125	7	33		Not ceiling-heavy under operational D frame; late/reach effects dominate.
year1	ASMC	30	0.476	0	0.011	0	0	0.133	6	24		Not ceiling-heavy under operational D frame; late/reach effects dominate.
year1	BNL0-100	13		0		0			13	0	0.264	No NL2 ceiling signal; median exact/near-exact category rate is not high.
year1	MC0-100	60	0.892	15	0.002	2	0	0.2	14	46		Small number of very easy D-frame items; no subtest-level ceiling.
year1	MNC0-100	29	0.650	2	0.021	1	0	0.207	8	21		Small number of very easy D-frame items; no subtest-level ceiling.

Excluded-from-fit item reasons

Most excluded timed math items are not too easy; they are late/unreached or all-zero under the D frame. DMT is the exception: excluded items were all correct.

year_level	test_subgroup	total_excluded_from_fit	no_valid_responses_late_unreached	all_valid_incorrect_or_late_all_zero	all_valid_correct_too_easy
foundation	DMT10_2026	3			3
foundation	MC0-20	38	4	16	18
foundation	MNC0-20	12	6	6
foundation	MQ1-20	24	16	6	2
year1	AAMC	13		4	9
year1	ASMC	10		10
year1	MC0-100	58	50	2	6
year1	MNC0-100	14	8	6

2025 CIS anchor context

The 2025 CIS scale file is a different artefact: scale lookups/anchors rather than the 2026 D-frame item audit. High positive cohort theta means and large negative b shifts in several 2025 accuracy forms are consistent with forms sitting below the cohort ability distribution.

year_level	subtest	form_test_id	kind	cohort_theta_mean	b_shift_applied	draft_for_revision	cohort_filter
foundation	BNL0-20	BNL0-20F_2025-MOY	placement	-0.0	0.0		ALS-fitted cohort (boundary-clipped at (0.001, 0.999); no person-level Fisher trim)
foundation	BNL0-20	BNL0-20F_2025	placement	0.0	-0.0		ALS-fitted cohort (boundary-clipped at (0.001, 0.999); no person-level Fisher trim)
foundation	DMT10	DMT10_2025	accuracy	-0.849	0.715		trimmed (Fisher info at hardest item < 0.10 AND theta > b_max; 77/1181 = 6.5% excluded)
foundation	MC0-20	MC0-20_2025	accuracy	2.746	-3.267		trimmed (Fisher info at hardest item < 0.10 AND theta > b_max; 0/4985 = 0.0% excluded)
foundation	MNC0-20	MNC0-20_2025	accuracy	1.578	-2.807	true	unfiltered (no Fisher filter computed for this subtest)
foundation	MQ1-20	MQ1-20_2025	accuracy	1.618	-2.029	true	unfiltered (no Fisher filter computed for this subtest)
year1	AAMC	AADD_2025-NEW	accuracy	2.66	-3.786	true	unfiltered (no Fisher filter computed for this subtest)
year1	AAMC	AAMC_2025-MOY	accuracy	2.569	-3.694	true	unfiltered (no Fisher filter computed for this subtest)
year1	ASMC	ASDD_2025_NEW	accuracy	1.378	-2.066		trimmed (Fisher info at hardest item < 0.10 AND theta > b_max; 0/1786 = 0.0% excluded)
year1	ASMC	ASMC_MOY	accuracy	1.725	-2.412		trimmed (Fisher info at hardest item < 0.10 AND theta > b_max; 0/1786 = 0.0% excluded)
year1	BNL0-100	BNL0-100_2025-MOY	placement	0.0	-0.0		ALS-fitted cohort (boundary-clipped at (0.001, 0.999); no person-level Fisher trim)
year1	BNL0-100	BNL0-100_2025	placement	-0.0	0.0		ALS-fitted cohort (boundary-clipped at (0.001, 0.999); no person-level Fisher trim)
year1	MC0-100	MC0-100_2025	accuracy	2.72	-3.229		trimmed (Fisher info at hardest item < 0.10 AND theta > b_max; 0/4968 = 0.0% excluded)
year1	MNC0-100	MNC0-100_2025	accuracy	1.941	-3.676	true	unfiltered (no Fisher filter computed for this subtest)

Data and scoring audit checks

These are the original ingestion, scoring, data-quality, chronology, Number Line calibration, item flag, and model-frame readiness sections from the audit.

Executive summary

996/1024

Foundation complete expected battery

1223/1256

Year 1 complete expected battery

997

foundation operational-frame persons
130259 binary rows; 9060 NL rows

1221

year1 operational-frame persons
192517 binary rows; 14734 NL rows

Decision posture: 2026 BOY appears to be one common fixed battery per year level, no explicit A/B assessment-event split. STPM is retained as shadow speed context only. DMT10_2026 is treated as untimed. Timed non-NL operational preparation uses D/trailing-zero; Number Line is recomputed from coordinates and encoded as NL2.

Data preservation and file-level QC

file_key	bytes	sha256	run_raw_path
foundation_csv	37769419	e16f3eb8217c889487ae4da344764c3096182e58738ba4c06dca6dafc074e2b1	outputs/runs/irt-2026-boy-subtest-audit/20260612T154956Z_2026_boy_subtest_item_audit/raw/Answer Report - Foundations BOY 2026.csv
foundation_parquet	500380	8730a878fd16a231bea48135f58964cd46e836ac61a2963f275008170dbc28d3	outputs/runs/irt-2026-boy-subtest-audit/20260612T154956Z_2026_boy_subtest_item_audit/raw/Answer Report - Foundations BOY 2026.parquet
year1_csv	57753657	9e0626be6882deecdafa79e2852673ce4b8dcaf028f6beea7b020a3be3e2a3ca	outputs/runs/irt-2026-boy-subtest-audit/20260612T154956Z_2026_boy_subtest_item_audit/raw/Answer Report - Year 1 BOY 2026.csv
year1_parquet	646713	fba3e842ad2e2c476c53e4467c3d87c46cd1816ebcfd10f493c2c614b134b0fa	outputs/runs/irt-2026-boy-subtest-audit/20260612T154956Z_2026_boy_subtest_item_audit/raw/Answer Report - Year 1 BOY 2026.parquet

source_file	rows	unique_students	practice_rows	missing_raw_score	missing_duration	blank_or_empty_array_response	placeholder_no_response	exam_groups
Answer Report - Foundations BOY 2026.csv	156786	1024	6018	2602	94530	1006	94317	ENSSA Foundation 2026 (1022 students); ENSSA Year 1 2026 (2 students)
Answer Report - Year 1 BOY 2026.csv	233919	1256	11072	5079	152572	1250	151942	ENSSA Year 1 2026 (1256 students)

No A/B cohort or assessment-event field is present in the supplied CSV schema. Cohort is coded as not_explicit. The Foundation file contains a small metadata anomaly: 2 students / 312 rows have ENSSA Year 1 2026 in Exam Group while taking Foundation forms; they are retained as Foundation-form data and flagged in the QC table.

Battery coverage and chronology

year_level	cohort	test_subgroup	test_group	is_timed	role	unique_students	n_items_nonpractice
foundation	not_explicit	STPM	speed_test	TRUE	non_math	1024	9
foundation	not_explicit	MQ1-20	match_quant	TRUE	achievement_primary	1006	30
foundation	not_explicit	MC0-20	mag_comp	TRUE	achievement_primary	1005	60
foundation	not_explicit	MNC0-20	missing	TRUE	achievement_primary	1003	30
foundation	not_explicit	DMT10_2026	decomp	FALSE	achievement_primary	1002	11
foundation	not_explicit	BNL0-20	numb_line	FALSE	achievement_primary	996	10
year1	not_explicit	STPM	speed_test	TRUE	non_math	1256	9
year1	not_explicit	MC0-100	mag_comp	TRUE	achievement_primary	1235	60
year1	not_explicit	MNC0-100	missing	TRUE	achievement_primary	1229	29
year1	not_explicit	AAMC	arithmetic	TRUE	achievement_primary	1227	40
year1	not_explicit	ASMC	arithmetic	TRUE	achievement_primary	1223	30
year1	not_explicit	BNL0-100	numb_line	FALSE	achievement_primary	1223	13

Observed fixed order

year_level	observed_order_by_median_start	test_subgroup	n_students_with_date	first_attempted_at	median_attempted_at	last_attempted_at
foundation	1	STPM	1001	2026-05-08 10:36	2026-06-09 12:44	2026-06-12 15:03
foundation	2	MQ1-20	998	2026-05-08 10:39	2026-06-09 12:47	2026-06-12 15:04
foundation	3	MC0-20	995	2026-05-08 10:45	2026-06-09 12:51	2026-06-12 15:06
foundation	4	MNC0-20	993	2026-05-08 10:50	2026-06-09 12:54	2026-06-12 15:10
foundation	5	DMT10_2026	988	2026-05-08 12:43	2026-06-09 12:56	2026-06-12 15:13
foundation	6	BNL0-20	979	2026-05-08 12:51	2026-06-09 13:01	2026-06-12 15:16
year1	1	STPM	1226	2026-05-07 09:35	2026-06-04 13:21	2026-06-12 15:15
year1	2	MC0-100	1221	2026-05-07 09:37	2026-06-04 13:24	2026-06-12 15:17
year1	3	MNC0-100	1211	2026-05-07 09:39	2026-06-04 14:30	2026-06-12 15:07
year1	4	AAMC	1205	2026-05-07 09:41	2026-06-04 14:58	2026-06-12 15:09
year1	5	ASMC	1199	2026-05-07 09:43	2026-06-05 12:27	2026-06-12 15:12
year1	6	BNL0-100	1180	2026-05-07 09:46	2026-06-09 09:12	2026-06-12 15:14

Battery patterns

year_level	subtest_pattern	n_students	is_complete_expected_battery	missing_expected_subtests
foundation	STPM + MQ1-20 + MC0-20 + MNC0-20 + DMT10_2026 + BNL0-20	996	TRUE
foundation	STPM	18	FALSE	MQ1-20;MC0-20;MNC0-20;DMT10_2026;BNL0-20
foundation	STPM + MQ1-20 + MC0-20 + MNC0-20 + DMT10_2026	6	FALSE	BNL0-20
foundation	STPM + MQ1-20 + MC0-20	2	FALSE	MNC0-20;DMT10_2026;BNL0-20
foundation	STPM + MQ1-20	1	FALSE	MC0-20;MNC0-20;DMT10_2026;BNL0-20
foundation	STPM + MQ1-20 + MC0-20 + MNC0-20	1	FALSE	DMT10_2026;BNL0-20
year1	STPM + MC0-100 + MNC0-100 + AAMC + ASMC + BNL0-100	1223	TRUE
year1	STPM	21	FALSE	MC0-100;MNC0-100;AAMC;ASMC;BNL0-100
year1	STPM + MC0-100	6	FALSE	MNC0-100;AAMC;ASMC;BNL0-100
year1	STPM + MC0-100 + MNC0-100 + AAMC	4	FALSE	ASMC;BNL0-100
year1	STPM + MC0-100 + MNC0-100	2	FALSE	AAMC;ASMC;BNL0-100

Response process and D/trailing-zero policy

year_level	test_subgroup	n_selected_nonpractice_rows	valid_scored_rows	trailing_nonresponse_rows	intermittent_nonresponse_rows	not_scored_contract_rows	missing_duration_rows	placeholder_no_response_rows	d_included_rows	d_trailing_zero_rows
foundation	STPM	9216	8637	480	0	0	576	576	0	0
foundation	MQ1-20	30180	4775	25135	0	0	25404	25404	29910	25135
foundation	MC0-20	60300	15067	44633	0	0	45233	45233	59700	44633
foundation	MNC0-20	30090	7188	22602	0	0	22901	22901	29790	22602
foundation	DMT10_2026	11022	10859	0	7	2	161	161	10859	0
foundation	BNL0-20	9960	9060	0	680	0	198	0	0	0
year1	STPM	11299	10570	500	0	0	724	724	0	0
year1	MC0-100	74100	16954	56305	0	1	57138	57138	73259	56305
year1	MNC0-100	35641	9654	25465	0	0	25981	25981	35119	25465
year1	AAMC	49080	9695	38505	0	0	39381	39381	48200	38505
year1	ASMC	36690	8204	27735	0	1	28480	28480	35939	27735
year1	BNL0-100	15899	14734	0	580	0	600	0	0	0

Timed non-NL rows after a student's last valid scored response are coded as trailing unreached and receive zero in the D frame. Intermittent/QC/contract missing rows remain missing. Untimed DMT and Number Line do not use trailing-zero scoring.

Subtest score and response-time summaries

year_level	test_subgroup	n_students	median_raw_sumscore_source	sd_raw_sumscore_source	median_d_or_valid_sumscore	sd_d_or_valid_sumscore	median_nl_mean_accuracy	median_total_rt_sec	sd_total_rt_sec	median_n_reached_or_valid_count	median_trailing_length
foundation	STPM	1024	9	1.887	9	1.887		88	146.682	9	0
foundation	MQ1-20	1006	3	1.648	3	1.648		105	17.112	4	26
foundation	MC0-20	1005	14	5.509	14	5.509		98	23.198	15	45
foundation	MNC0-20	1003	5	3.339	5	3.339		105	14.602	7	23
foundation	DMT10_2026	1002	8	2.073	8	2.073		238	220.724	11	0
foundation	BNL0-20	996	1	1.155			0.873	83	63.975	10	0
year1	STPM	1256	9	2.04	9	2.04		78	61.495	9	0
year1	MC0-100	1235	12	5.762	12	5.762		100	12.776	14	46
year1	MNC0-100	1229	6	3.347	6	3.347		105	25.362	8	21
year1	AAMC	1227	5	4.141	5	4.141		105	19.386	7	33
year1	ASMC	1223	4	3.16	4	3.16		105	14.824	6	24
year1	BNL0-100	1223	0	0.593			0.866	90	60.704	13	0

Number Line coordinate audit

year_level	test_subgroup	n_coordinate_rows	x_min	x_p01	x_median	x_p99	x_max	chosen_x_l	chosen_x_r	n_out_of_bounds_chosen	out_of_bounds_rate_chosen	correct_click_implied_x_at_vmax	moy_residual_rmse_px	boy_residual_rmse_px	calibration_decision
foundation	BNL0-20	9060	-1	16.094	489.58	1107	1129	8.5	1108	125	0.014	1108.425	15.224	16.003	use_2025_MOY_endpoints; correct-click and out-of-range diagnostics support reuse
year1	BNL0-100	14734	0	38.33	503	1082	1117	10.5	1106	51	0.003	1104.211	3.829	9.791	use_2025_MOY_endpoints; correct-click and out-of-range diagnostics support reuse

The answer payloads are JSON-like ValuePoint.X/Y coordinates. The audit uses 2025 MOY bounded Number Line endpoints: BNL0-20 x_l=8.5, x_r=1108; BNL0-100 x_l=10.5, x_r=1106. Platform Number Line raw score is not used for operational scoring.

Item flags

year_level	test_subgroup	n_items	n_items_with_any_flag	flag_breakdown
foundation	BNL0-20	10	1	high_missing_coordinate_rate=1
foundation	DMT10_2026	11	3	very_high_p_correct_valid=3
foundation	MC0-20	60	52	high_trailing_rate=49; very_high_p_correct_valid=22; very_low_p_correct_valid=8
foundation	MNC0-20	30	25	high_trailing_rate=25; low_item_total_proxy=5; very_low_p_correct_valid=4
foundation	MQ1-20	30	27	high_trailing_rate=26; low_item_total_proxy=1; very_high_p_correct_valid=2; very_low_p_correct_valid=3
foundation	STPM	9	9	very_high_p_correct_valid=9
year1	AAMC	40	35	high_trailing_rate=35; very_high_p_correct_valid=9; very_low_p_correct_valid=3
year1	ASMC	30	26	high_trailing_rate=26; low_item_total_proxy=4; very_low_p_correct_valid=5
year1	BNL0-100	13	0
year1	MC0-100	60	53	high_trailing_rate=51; very_high_p_correct_valid=8; very_low_p_correct_valid=1
year1	MNC0-100	29	24	high_trailing_rate=24; low_item_total_proxy=4; very_low_p_correct_valid=3
year1	STPM	9	9	very_high_p_correct_valid=9

Standalone subtest IRT diagnostics

year_level	test_subgroup	policy_variant	n_persons	n_items	n_complete_cases_for_alpha	alpha_or_kr20_proxy_complete_case	median_proxy_score	sd_proxy_score	diagnostic_model_status
foundation	STPM	D trailing-zero	1024	9	1013	0.888	9	1.689	empirical_proxy_only_local_run
foundation	MQ1-20	D trailing-zero	1006	30	997	0.582	3	1.63	empirical_proxy_only_local_run
foundation	MC0-20	D trailing-zero	1005	60	995	0.896	14	5.367	empirical_proxy_only_local_run
foundation	MNC0-20	D trailing-zero	1003	30	993	0.83	5	3.315	empirical_proxy_only_local_run
foundation	DMT10_2026	valid-only	1002	11	986	0.584	8	1.868	empirical_proxy_only_local_run
foundation	BNL0-20	NL2 ordinal	996	10	633	0.704	9	4.063	empirical_proxy_only_local_run
year1	STPM	D trailing-zero	1256	9	1230	0.915	9	1.665	empirical_proxy_only_local_run
year1	MC0-100	D trailing-zero	1235	60	1220	0.902	12	5.647	empirical_proxy_only_local_run
year1	MNC0-100	D trailing-zero	1229	29	1211	0.839	6	3.281	empirical_proxy_only_local_run
year1	AAMC	D trailing-zero	1227	40	1205	0.87	5	4.095	empirical_proxy_only_local_run
year1	ASMC	D trailing-zero	1223	30	1197	0.816	4	3.138	empirical_proxy_only_local_run
year1	BNL0-100	NL2 ordinal	1223	13	877	0.743	11	5.006	empirical_proxy_only_local_run

The first table is a dependency-light empirical proxy retained for reproducibility. The package-backed TAM table below was fit on cisbox from the same classified rows. The operational Stan JSONs are built but not fit by this audit script.

cisbox package-backed standalone TAM fits

year_level	test_subgroup	policy_variant	n_persons_matrix	n_items_matrix	n_removed_constant_or_empty_items	tam_status	tam_irtmodel	eap_reliability	n_iterations
foundation	STPM	D trailing-zero Rasch 1PL	1013	9	0	fit_ok	1PL	0.524063471225759	41
foundation	MQ1-20	D trailing-zero Rasch 1PL	997	19	11	fit_ok	1PL	0.602083755686912	32
foundation	MC0-20	D trailing-zero Rasch 1PL	995	50	10	fit_ok	1PL	0.92686212002155	68
foundation	MNC0-20	D trailing-zero Rasch 1PL	993	24	6	fit_ok	1PL	0.88127137044596	36
foundation	DMT10_2026	valid-only Rasch 1PL	988	8	3	fit_ok	1PL	0.609334468054377	26
foundation	BNL0-20	NL2 ordinal PCM	974	10	0	fit_ok	PCM	0.674311095384489	21
year1	STPM	D trailing-zero Rasch 1PL	1230	9	0	fit_ok	1PL	0.470479868039455	36
year1	MC0-100	D trailing-zero Rasch 1PL	1221	34	26	fit_ok	1PL	0.940382239713017	77
year1	MNC0-100	D trailing-zero Rasch 1PL	1211	22	7	fit_ok	1PL	0.891226639916213	45
year1	AAMC	D trailing-zero Rasch 1PL	1205	38	2	fit_ok	1PL	0.900370448060522	44
year1	ASMC	D trailing-zero Rasch 1PL	1198	25	5	fit_ok	1PL	0.841455947796902	40
year1	BNL0-100	NL2 ordinal PCM	1178	13	0	fit_ok	PCM	0.727239296568283	17

TAM standalone subtest models were fit on cisbox after local intake: binary/timed subtests use the D trailing-zero Rasch 1PL frame, DMT10_2026 uses valid-only Rasch 1PL, and Number Line uses NL2 ordinal PCM. These are diagnostic fits only, not the operational D-A2/NL2 global model.

2025 T3/T4 form continuity

year_level	test_subgroup	in_2026_boy	in_2025_t3	in_2025_t4	overlap_note
foundation	STPM	TRUE	TRUE	TRUE	retained exact canonical subtest
foundation	MQ1-20	TRUE	TRUE	TRUE	retained exact canonical subtest
foundation	MC0-20	TRUE	TRUE	TRUE	retained exact canonical subtest
foundation	MNC0-20	TRUE	TRUE	TRUE	retained exact canonical subtest
foundation	DMT10_2026	TRUE	FALSE	FALSE	new 2026 decomposition form; family-level overlap with 2025 DMT5V2/DMT10V2/DMT10 only
foundation	BNL0-20	TRUE	TRUE	TRUE	retained exact canonical subtest
foundation	DMT10V2	FALSE	TRUE	TRUE	absent in 2026
foundation	DMT5V2	FALSE	TRUE	TRUE	absent in 2026
foundation	MNA0-20	FALSE	TRUE	TRUE	absent in 2026
foundation	MQ1-10	FALSE	TRUE	TRUE	absent in 2026
foundation	UNLC0-20	FALSE	FALSE	TRUE	absent in 2026
foundation	UNLNC0-20	FALSE	TRUE	TRUE	absent in 2026
year1	STPM	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	MC0-100	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	MNC0-100	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	AAMC	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	ASMC	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	BNL0-100	TRUE	TRUE	TRUE	retained exact canonical subtest
year1	ASDD	FALSE	TRUE	FALSE	absent in 2026
year1	MNA0-100	FALSE	TRUE	TRUE	absent in 2026
year1	STDD	FALSE	TRUE	FALSE	absent in 2026
year1	UNLC0-20	FALSE	TRUE	TRUE	absent in 2026
year1	UNLNC0-20	FALSE	TRUE	TRUE	absent in 2026

Operational D-A2/NL2 model-frame readiness

year_level	model_key	N_person	N_testlets	N_acc_items	N_acc_bin	N_nl_items	N_acc_nl	achievement_subtests	excluded_subtests	ready_for_stan_data_smoke	fit_status
foundation	D_A2NL2_BOY2026_operational_candidate__foundation	997	5	141	130259	10	9060	MQ1-20;MC0-20;MNC0-20;DMT10_2026;BNL0-20	STPM	TRUE	not_fit_by_audit_script
year1	D_A2NL2_BOY2026_operational_candidate__year1	1221	5	172	192517	13	14734	MC0-100;MNC0-100;AAMC;ASMC;BNL0-100	STPM	TRUE	not_fit_by_audit_script

Not a promotion decision: these JSONs are data-frame inputs only. Full model fitting, diagnostics, validation, subgroup/fairness audit, band calibration, and governance sign-off remain separate gates.

Diagnostic figures

score_histogram: foundation BNL0-20

score_histogram: foundation DMT10_2026

score_histogram: foundation MC0-20

score_histogram: foundation MNC0-20

score_histogram: foundation MQ1-20

score_histogram: foundation STPM

score_histogram: year1 AAMC

score_histogram: year1 ASMC

score_histogram: year1 BNL0-100

score_histogram: year1 MC0-100

score_histogram: year1 MNC0-100

score_histogram: year1 STPM

reached_count_histogram: foundation BNL0-20

reached_count_histogram: foundation DMT10_2026

reached_count_histogram: foundation MC0-20

reached_count_histogram: foundation MNC0-20

reached_count_histogram: foundation MQ1-20

reached_count_histogram: foundation STPM

reached_count_histogram: year1 AAMC

reached_count_histogram: year1 ASMC

reached_count_histogram: year1 BNL0-100

reached_count_histogram: year1 MC0-100

reached_count_histogram: year1 MNC0-100

reached_count_histogram: year1 STPM

rt_distribution: foundation BNL0-20

rt_distribution: foundation DMT10_2026

rt_distribution: foundation MC0-20

rt_distribution: foundation MNC0-20

rt_distribution: foundation MQ1-20

rt_distribution: foundation STPM

rt_distribution: year1 AAMC

rt_distribution: year1 ASMC

rt_distribution: year1 BNL0-100

rt_distribution: year1 MC0-100

rt_distribution: year1 MNC0-100

rt_distribution: year1 STPM

item_curve: foundation STPM

item_curve: foundation MQ1-20

item_curve: foundation MC0-20

item_curve: foundation MNC0-20

Artifact index

Internal/person-level artifacts contain student identifiers and should remain in the ignored local run root and cisbox artifact store.

tables/2026_boy_row_qc.csv
tables/2026_boy_subtest_mapping.csv
tables/2026_boy_unique_students_by_subtest.csv
tables/2026_boy_student_battery_patterns.csv
tables/2026_boy_subtest_score_rt_summary.csv
tables/2026_boy_response_process_by_subtest.csv
tables/2026_boy_item_level_audit.csv
tables/2026_boy_number_line_coordinate_audit.csv
tables/2026_boy_overlap_vs_2025_t3_t4.csv
tables/2026_boy_operational_model_frame_manifest.csv
model_data/foundation/D_A2NL2_BOY2026_operational_candidate__foundation.json
model_data/year1/D_A2NL2_BOY2026_operational_candidate__year1.json

Choose a Foundation subtest

STPM Non-math; not in math score

TimedNon-mathPrimary policyD/trailing-zeroReliability0.524Excluded0

Model / scoring policy view

Decision: Timed non-math subtest; retained for non-math diagnostics, not included in the math achievement score.

Policy table below lets you compare the operational-candidate view with diagnostic policy alternatives where available.

year_level	test_subgroup	timing	primary_policy	decision_basis	eap_reliability	excluded_from_fit_rate	itemfit_outside_0.75_1.25_rate
foundation	STPM	Timed	D/trailing-zero Rasch 1PL	Subtest is STPM/non-math. It receives timed policy diagnostics but is excluded from the math achievement score.	0.524	0.000	0.889

Policy comparison

policy	use	EAP reliability	fit items	excluded from fit	excluded rate	status
D/trailing-zero	Operational candidate	0.524	9	0	0.000	fit_ok
Reached/valid-only	Diagnostic/comparability	0.252	9	0	0.000	fit_ok

D/trailing-zero — Wright map

Reached/valid-only — Wright map

D/trailing-zero — Item fit

item	Infit	Outfit	flag
STPM010	1.67	4.48	outside 0.75–1.25
STPM005-copy-3	0.55	0.14	outside 0.75–1.25
STPM006-copy-3	0.56	0.15	outside 0.75–1.25
STPM002-copy-3	0.81	0.17	outside 0.75–1.25
STPM007-copy-3	0.57	0.19	outside 0.75–1.25
STPM008-copy-3	0.64	0.28	outside 0.75–1.25
STPM009-copy-3	0.76	0.59	outside 0.75–1.25
STPM004-copy-3	0.73	0.62	outside 0.75–1.25
STPM003-copy-3	0.95	0.98	within band

Reached/valid-only — Item fit

item	Infit	Outfit	flag
STPM005-copy-3	0.82	0.44	outside 0.75–1.25
STPM002-copy-3	0.86	0.49	outside 0.75–1.25
STPM006-copy-3	0.87	0.72	outside 0.75–1.25
STPM004-copy-3	0.99	1.25	outside 0.75–1.25
STPM008-copy-3	1.07	1.25	within band
STPM003-copy-3	1.01	1.24	within band
STPM009-copy-3	0.99	0.87	within band
STPM010	1.05	1.09	within band
STPM007-copy-3	0.91	1.04	within band

D/trailing-zero — TCC

Reached/valid-only — TCC

D/trailing-zero — item curves

Item curve

STPM010

Reached/valid-only — item curves

Item curve

STPM010