# Data Science 301 — Project report

**Question:** Are male and female A-level Maths entries in England in 2024/25 associated
with different A*–B grade outcomes?

## 1. Introduction — why this hypothesis

In Data Science 201 I tested whether Computer Science **entries** are gender-balanced.
They are not. Maths is the largest individual A-level subject and is also male-skewed in
volume, so a natural follow-on is whether **grade outcomes** differ by sex too. If A*–B
rates diverged sharply, that would be a different equity story from “who enters.” If they
barely differ, the imbalance is mainly about participation, not measured attainment among
those who enter.

## 2. Human question

Among A-level Maths entries in England in 2024/25, do female and male students achieve
A*–B at different rates?

## 3. Hypothesis

The A*–B achievement rate for female Maths entries differs from the rate for male Maths
entries in England in 2024/25.

## 4. Null hypothesis

Female and male Maths entries have the same underlying A*–B achievement probability.

## 5. Required fields

`subject_name`, `time_period`, `country_name`, `geographic_level`,
`characteristic_value`, `entry_count`, `perc_astar_b_grade_achieved`.

## 6. Data cleaning

- Filter to Maths, England, National, 2024/25, Female / Male / All students.
- Convert `perc_astar_b_grade_achieved` to numeric.
- Reconstruct approximate A*–B counts as `round(entry_count × percentage / 100)` because
  the public file publishes percentages, not raw grade tallies.
- Note: Female + Male = **101,066**, All students = **101,079**
  (difference **13**), consistent with DfE footnotes that sex is not always recorded.

## 7. Statistical test

Two-sided **two-proportion z-test** on reconstructed A*–B counts:

| Group | Entries | A*–B % | Approx. A*–B count |
| --- | ---: | ---: | ---: |
| Female | 37,612 | 61.8% | 23,244 |
| Male | 63,454 | 62.5% | 39,659 |

- Difference (female − male): **-0.70** percentage points  
- z ≈ **-2.22**, p ≈ **0.0263**

## 8. Chart

`outputs/charts/ds301_maths_volume_vs_grades.png`

## 9. Conclusion — supported or rejected?

**Statistically:** the evidence **supports** the hypothesis that the female and male A*–B
rates are not identical. At conventional α = 0.05 we **reject the null** (p ≈ 0.026).

**Practically:** the hypothesis is only weakly interesting once you see the size of the gap.
Female Maths entries achieve A*–B at **61.8%** and male entries at
**62.5%** — about **0.7** percentage points apart. With
n > 100,000 that tiny gap is detectable, but it does **not** look like a meaningful
attainment divide.

**Verdict for a client:** I would **not** claim a material sex difference in Maths A*–B
outcomes. Entries are substantially male-skewed; **grade rates among entrants are nearly the
same**. That contrasts with Computer Science in Data Science 201, where the entry split
itself was far from 50/50.

## 10. Limitations

1. A*–B counts are **reconstructed from rounded percentages**, not official grade tallies.
2. Results are **entries**, not unique students; students may take multiple subjects.
3. National aggregate only — no school, region or prior-attainment controls.
4. A statistically significant p-value with n > 100,000 does not imply an important gap.
5. No causal claim about teaching, confidence or subject choice.

## Reflection

**Hardest part:** Choosing a question narrow enough to finish, after noticing how easy it
would be to re-run the Computer Science gender test on every STEM subject. **Most enjoyable:**
Finding that volume imbalance and grade outcomes can tell different stories in the same
subject. I would next bring in prior-attainment or disadvantage breakdowns from a wider DfE
file if the client cared about equity of *outcomes* rather than *entries* alone.