How often each topic actually gets tested

Confirmed tested, with at least one full question devoted to it, in 5 of the 6 series with a fully-read examiner report (Jan 2021, Oct 2021, Jun 2022, Jan 2023, Jun 2024) — the richest single sub-topic in the whole facts bank, and the single largest question on the one series with a full mark-weight reconstruction (17 of 75 marks, Jan 2023). The conditional-probability sub-part, when present, is consistently the last lettered part of its question and is repeatedly flagged as the paper's most discriminating content.

Confirmed in 5 of the 6 series with a fully-read examiner report (Jan 2021, Oct 2021, Jun 2022, Jun 2024, Jun 2019) as its own tested skill, distinct from histogram-reading, mean/SD or coding. Never a standalone question — always a part-set inside a larger opening Section-2 question — and three genuinely distinct angles recur: fence-rule application, box-plot/whisker construction, and comparing two distributions in words.

Confirmed in 4 of the 6 series with a fully-read examiner report (Jan 2021, Oct 2021, Jun 2022, Jan 2023) — independence and mutual exclusivity treated as a calculation, not a judgment. The more damaging of two recurring angles is a silent assumption buried inside an unrelated calculation, which Pearson's own Jun 2022 report names as a paper-wide diagnosis, not a one-question slip.

Confirmed in 4 of the 6 series with a fully-read examiner report (Jan 2021, Oct 2021, Jun 2022, Jan 2023) — this topic's own notation/counting error, distinct from the sibling independence-judgment brief. Spec 3.1 alone has no standalone question anywhere in the 14-series archive; every citation is really a 3.1+3.2 composite, and the recurring failure is miscounting exactly which terms belong in a conditional probability's numerator.

Confirmed in 4 of the 6 series with a fully-read examiner report (Jan 2021, Oct 2021, Jun 2022, Jan 2023) for this lesson's own core — narrower than the paper-wide Section-2 'confirmed every series' figure, which pools in histogram-reading and outliers. The one thread that repeats across series, not just within one, is a squaring step in variance/SD arithmetic dropped or mishandled.

Confirmed in 4 of the 6 series with a fully-read examiner report (Jan 2023, Jun 2022, Oct 2021, Jun 2024), splitting into two sub-skills that don't always co-occur: reading a histogram or stem-and-leaf diagram (2 of 6) and comparing two distributions in words (3 of 6). The comparing-in-words step is never standalone — always the final sub-part of whichever data-representation question that series builds — and most often opens the paper as part of Q1.

Confirmed in 4 of the 6 series with a fully-read examiner report (Oct 2021, Jun 2022, Jan 2023, Jun 2024), always the back half of a larger regression question, never its own. Two failure families recur: reading the gradient at the wrong scale of the explanatory variable, and a bare direction-only sentence with no rate of change attached — a reliability verdict with no named reference point ('extrapolation' alone) is confirmed worthless in two separate series.

Confirmed in 3 of the 6 series with a fully-read examiner report (Oct 2021, Jun 2022, Jun 2024) — the PMCC calculation itself, correcting an earlier synthesis that had missed the Jun 2022 instance. Never a standalone question; always one early-middle part of a larger correlation-and-regression question, with at least one sum already given rather than built from raw data unaided.

Confirmed in 3 of the 6 series with a fully-read examiner report (Oct 2021, Jan 2023, Jun 2024) — stricter than one existing synthesis's '4 of 6,' which counts a converse, non-'show that' case rather than a real instance. Cross-cutting by spec section (confirmed in Sections 2, 4, 5 and 6; no confirmed instance in Section 3); risk concentrates in the final mark of the part, an A1cso or A1*, not spread evenly.

Narrowed to this lesson's own scope (excluding the E(aX+b)/Var(aX+b) sibling topic), confirmed in 2 of the 6 series with a fully-read examiner report by direct quote (Oct 2021, Jun 2022), plus a 1-mark mark-scheme-only instance (Jun 2024) and one series outside the 6-series base entirely (Jan 2025, where it carried a full opening question worth 15 of 75 marks). The discrete uniform distribution has never anchored a multi-part question by itself in the reviewed record.

Confirmed in 2 of the 6 series with a fully-read examiner report (Jan 2021, Jun 2024 — Jun 2024 contributing two separate questions), narrower than a facts-bank summary line that rounds this to '3 separate series.' The single most-repeated error, three times over two series, is dropping the square in Var(aX+b)=a²Var(X); both real anchors are unsolvable unless the variance equation is solved before the mean equation.

No standalone question testing mathematical modelling in isolation exists anywhere in the 14-series archive — a confirmed absence, not an unchecked gap. The concept surfaces only as a single opening B1 mark tied to a stated assumption (Jan 2025 Q1(a)) or as a silent, embedded trap inside a probability or regression-reliability question (both instances Jun 2022) — every credited answer is specific (a named distribution family, or a named value and range), never a generic qualitative statement.

This lesson has no dedicated topic brief in this course's research, unlike every other WST01 lesson — its content is folded into the paper-wide table's pooled 'Probability, 3.1–3.4: seen in 4 of 6' row alongside conditional probability and independence, and cannot be honestly isolated from that pooled figure with what this course has verified so far. Too thin a base to tier on its own evidence; treat this row as a flagged gap in the research, not a frequency finding.