Technique Masterclasses

Statistics 1

Topic Brief: Mathematical Modelling in Probability and Statistics

SIGNAL Topic Master Brief — Mathematical Modelling in Probability and Statistics (WST01, spec 1.1)

Provenance (2026-09-29): built entirely from research/veridian/WST01-verified-facts.md (compiled 2026-09-12, spot-check addendum 2026-09-25), grepped for "model", "modelling", "assum", "refin", "extrapolat", "reliab". Cross-checked against signal-frequency-and-boundaries.md (this folder, row 1.1) and codex-deep-dive-mathematical-modelling-in-probability-and-statistics.md (this folder, which already establishes the same underlying frequency counts through its own case analysis) so this document does not re-derive either — it repackages the confirmed evidence into the frequency / mark-tariff / context-angle / expect-summary shape this course's SIGNAL documents use, and adds one synthesis neither existing document states explicitly as its own subject: the mark- tariff pattern in §3. The live lesson (content/veridian/curriculum/wst01/mathematical-modelling-in-probability-and-statistics.ts) and content/veridian/question-bank/wst01.ts (checked — no item on this topic) were also consulted. No new PDF fetches were made for this document.

1. Frequency — seen in 0 of 14 series as a standalone question

The facts bank searched all 14 reviewed WST01 question papers (Jun 2019 → Jan 2025) directly for "model"/"modelling." Every hit resolved to a regression model reference (spec 4.1/4.2), not spec 1.1. No standalone question testing "the basic ideas of mathematical modelling" in isolation exists anywhere in this 14-series archive — a confirmed absence, not an unchecked gap.

Within that same 14-series archive, exactly one series carries a mark-scheme line that directly credits the spec-1.1 concept itself (naming the distributional family a stated assumption licenses): Jan 2025 Q1(a). Two further instances, both drawn from the 6 series with a fully-read examiner report (Jun 2019, Jan 2021, Oct 2021, Jun 2022, Jan 2023, Jun 2024), show the concept surfacing as an embedded failure mode inside a different spec item's question rather than as credited content of its own — both 1 of 6, and both from the same series (Jun 2022, two different questions).

2. What was actually tested, by series

3. Mark-tariff pattern

Every confirmed mark in this archive tied specifically to the modelling concept — as opposed to marks merely lost through an unlicensed modelling failure — is worth exactly one mark (B1): Jan 2025 Q1(a)'s "Discrete uniform," and Jun 2022 Q2(e)(ii)'s reliability judgement. Both share a second feature: the mark requires a specific piece of stated evidence (the named distribution family; the named value/variable outside the fitted range), not a generic assertion — a bare "unreliable," or an unattached claim about a model's assumptions, earns nothing in either citation. There is no confirmed instance anywhere in the 14-series archive of a multi-mark, standalone modelling question. The tariff, whenever it is confirmed at all, is a single B-mark riding on a larger question built around a different spec item.

4. Recurring context/angle across the confirmed instances

Two distinct angles, and only two, are confirmed:

  1. A licensed assumption, stated by the question, credited for naming the family it implies — Jan 2025's opening sub-part. The assumption itself earns nothing; the named consequence (the distribution) does.
  2. An assumption never licensed by the question, self-imposed by the candidate — Jun 2022 Q4(b)'s independence substitution, and, in a different guise, Jun 2022 Q2(e)(ii)'s regression-validity breach, where the candidate implicitly treats a fitted line as valid outside the range it was fitted to. Both are refinement failures in spec 1.1's own sense — an assumption that should have been questioned went unquestioned, or was invented from nothing — even though neither sits inside a question labelled "modelling."

5. If this topic appears, here's what to expect

Do not expect a standalone question naming "mathematical modelling" — none exists in the confirmed 14-series record. Expect the concept instead as a single opening B1 mark riding on a probability question that states an assumption outright (name the distribution family it implies, and nothing more elaborate), or as a silent, embedded trap inside a conditional-probability or regression- reliability question, where the actual risk is substituting an unlicensed assumption for the calculation the question requires, or asserting a model's validity outside the range it was fitted to without naming the specific value that breaches it. In either embedded case, the credited answer form is always specific — a named family, or a named value and range — never a generic qualitative statement.

Sources and limitations