Python Analytics

Nafs Results Analytics — Aseer Region

A complete Python pipeline that starts from the raw extract, applies data-quality rules and quarantines failures, then builds an analytical model of trends, gaps and school segments that feeds this executive dashboard.

1Raw extract70,041 rows
2DQ rulesDQ-001…007
3Modelpandas · SciPy
4DashboardJSON → Chart.js

Filters Year Grade Subject Gender Authority

Proficiency trend vs national

National benchmark weighted by the selected grade/subject mix

Key findings

Generated by the pipeline

Regional proficiency rose from 32.0% (1444 AH) to 40.4% (1447 AH), +8.4 pts at 2.9 pts per year.
Gap to the national average is -1.8 pts in 1447 AH vs -3.0 in 1444 AH.
Science leads (52.2%) and Mathematics trails (31.5%), a 20.7-pt spread; weakest sub-domain: Algebra (26.9%).
Top governorate Abha (45.3%), lowest Al-Harjah (29.5%); most improved Tanomah (+5.0).
Girls' schools outperform by 5.5 pts on average; the difference is statistically significant (Welch t-test, p < 0.001).
189 schools flagged for priority intervention: ≥5 pts below the regional average and down ≥5 pts year-on-year (of 2,137 classified).
On the current trend, 1448 AH proficiency is projected at ~43.4% (80% range: 42.5–44.2%).

Proficiency by subject & grade

Performance-level distribution

Share of students per level

Governorate ranking

Dashed line: regional average

Gender & authority

Sub-domain diagnosis

Grade 6 science & Grade 9 maths

School segmentation: level vs change

Each dot is a school · 1447 AH vs 1446 · min. 15 tested

Priority schools for intervention

≥5 pts below average and down ≥5 pts · pseudonymised IDs

School	Governorate	1447	Δ
`SCH-1727` · Girls	Ahad Rufaidah	8.9%	-22.4
`SCH-0770` · Boys	Bisha	16.7%	-23.9
`SCH-1801` · Boys	Bariq	11.1%	-16.2
`SCH-1822` · Girls	Al-Majardah	20.7%	-24.1
`SCH-1592` · Boys	Balqarn	18.5%	-21.5
`SCH-1631` · Boys	Balqarn	10.3%	-12.7
`SCH-1337` · Girls	Tathlith	5.9%	-7.9
`SCH-2114` · Boys	Al-Harjah	8.0%	-9.9
`SCH-0016` · Boys	Khamis Mushait	9.1%	-10.9
`SCH-1020` · Girls	Muhayil	6.9%	-7.4
`SCH-1684` · Girls	Ahad Rufaidah	14.8%	-15.2
`SCH-1625` · Boys	Balqarn	14.8%	-14.8

Total flagged: 189 · Full list in priority_schools.csv

Data quality before analysis

Raw-extract quality scorecard

The extract is checked against seven rules mapped to DAMA dimensions before any calculation. Failures are quarantined or conformed from reference data, with the action documented.

70,041

Raw rows

70,023

Rows certified

Quarantined

Duplicates removed

Values conformed

Completeness 99.99%

Consistency 100.00%

Validity 100.00%

Uniqueness 99.99%

Timeliness 100.00%

Scale 95–100% · dark tick = target

Rule	Dimension	Check	Checked	Failed	Pass	Action
`DQ-001`	Completeness	School ID is not null	70,041	5	99.99%	Quarantine & return to source
`DQ-002`	Consistency	Governorate matches reference data	70,041	2	100.00%	Auto-conform from reference table
`DQ-003`	Validity	Grade & domain codes are in reference lists	70,041	1	100.00%	Quarantine & fix coding
`DQ-004`	Validity	Proficiency within 0-100 and consistent with counts	70,041	6	99.99%	Recompute from source counts
`DQ-005`	Validity	Tested does not exceed expected	70,041	3	100.00%	Quarantine for source verification
`DQ-006`	Uniqueness	No duplicate year/school/grade/domain	70,041	9	99.99%	Remove duplicates after review
`DQ-007`	Timeliness	Extract covers all reporting years	4	0	100.00%	Escalate delay to data owner

Method & statistics

Behind the numbers

Girls vs boys schools+5.5 pts Welch t = 8.58 · p < 0.001 · n = 1,112 / 1,025

Private & intl. vs public+4.6 pts Welch t = 4.07 · p < 0.001

School size vs proficiencyρ = -0.005 Spearman · p = 0.817 — no material relationship

Annual improvement+2.9 pts/yr OLS · R² = 0.996

Projection for 1448 AH43.4% 80% prediction interval: 42.5–44.2%

Pipeline excerpt

nafs_pipeline.py

# DQ-004 validity — proficiency within 0–100 and consistent with counts
recomputed = 100 * df["proficient"] / df["tested"]
bad_rate = (df["prof_rate"] < 0) | (df["prof_rate"] > 100) \
           | ((recomputed - df["prof_rate"]).abs() > 0.5)
rule("DQ-004", "Validity", bad_rate.sum(), action="Recompute from source counts")

# Priority = materially below average AND materially declining
wide["priority"] = ((wide["rate_1447"] <= region_avg - MATERIAL)
                    & (wide["change"] <= -MATERIAL)
                    & (wide["tested_1447"] >= PRIORITY_N))

# Significance: girls' vs boys' schools (Welch t-test)
t = stats.ttest_ind(girls, boys, equal_var=False)

DataSchool-level extract in the Nafs report layout: year, grade, domain & sub-domain, tested, proficient, performance levels and national average (1444–1447 AH).
PrivacySchool identifiers are pseudonymised; school-level indicators are suppressed below 15 tested students.
MeasuresProficiency = proficient ÷ tested (weighted by tested), because the score scale changed across years.

View notebook nafs_analysis.ipynb nafs_pipeline.py

Static Python outputs for reports

Data note: figures on this page are generated from a modelled dataset that follows the Nafs report layout; the same pipeline runs unchanged on the official extract.