King's College London · FACULTY OF STATISTICS

7SSGN110 Chap.7 Correlation and Ranked Relationships

- one subject, every graph, every model, every mark
7 Chapters4-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 7 of 8 · 7SSGN110

Correlation and Ranked Relationships

Association starts with paired geometry

A correlation coefficient summarises how two variables co-vary across matched observations. Before calculating it, draw the paired scatterplot and confirm that every x value remains attached to the y value from the same site, subject or occasion.

Sorting columns independently or combining records with different identifiers can produce a precise number for a relationship that never existed in the data.

Use Pearson for a linear summary

Pearson's r divides covariance by the sample spread of both variables, producing a unit-free value between minus one and one. The sign gives direction and the magnitude describes linear strength.

A value near zero means little linear association, not necessarily no relationship. Curvature, clusters and opposing segments can cancel, while one influential pair can control the result.

Plotting is therefore part of the calculation rather than an optional illustration.

Use ranks when order is the defensible scale

Spearman's rho applies correlation logic to ranked values and can describe monotonic association when numerical gaps are not meaningful. Kendall's tau compares concordant and discordant pair orderings and offers another route for ranked data. Ties matter for both.

A ranked coefficient does not estimate a fixed raw-unit change, so interpretation should stay with order and monotonic direction rather than borrowing the language of a regression slope.

Separate strength, uncertainty and importance

A strong sample coefficient can be unstable when few complete pairs are available; a weak coefficient can be statistically detectable with many observations.

Report the coefficient, complete-pair count, uncertainty or p-value and the observed variable ranges. Sensitivity checks can show how an influential site changes the number, but deletion needs a documented data-quality or population reason. Numerical strength alone does not establish environmental importance.

Keep causal mechanism outside the coefficient

Correlation describes sample co-variation.

Elevation, season, measurement effort or habitat quality may influence both recorded variables, and the coefficient cannot select among those mechanisms. A defensible discussion names plausible confounders and the additional design or observation that would distinguish explanations. Prediction requires a fitted model and diagnostics; causation requires stronger design evidence.

Keep the statistical description and scientific interpretation in separate layers.

Respect range and subgroup structure

A coefficient calculated across several habitats can reflect separation between group means even when little association exists within each habitat. Colour or facet the scatterplot by scientifically relevant groups and compare the pooled story with within-group patterns.

Restriction to a narrow measurement range can weaken a coefficient without changing the underlying mechanism. Report the sampled ranges and avoid comparing coefficients from different ranges as though magnitude alone established a stronger process.

For revision, sketch linear, curved, clustered and influential-point patterns that could share a similar coefficient.

Explain what the number preserves and what the graph reveals. Then choose Pearson or a ranked measure from the measurement scale and geometry, finishing with one explicit causal limitation.

In this chapter

What this chapter covers

  • 01

    Covariance

  • 02

    Pearson's r

  • 03

    Scatterplot

  • 04

    Spearman's rho

  • 05

    Kendall's tau

  • 06

    Linearity

  • 07

    Confounding

Worked example · free

Calculate Pearson's r

Q [9 marks]. AskSia-authored practice. The marks shown support revision and are not an official university marking scheme. Centred x values are -3, -1, 1, 3 and centred y values are -4, 0, 1, 3. Calculate Pearson's correlation and qualify the result.
  • 3Calculate the sum of cross-products as 22.
  • 3Calculate the square-root denominator from 20 and 26.
  • 3Interpret strength, direction and sample-size limits.
r = 22 divided by the square root of 20 times 26 = 22/22.80 = 0.965. The four sites show a strong positive linear association, but the estimate is unstable and does not establish a causal mechanism.
Sia tip — Check the scatterplot and the meaning of larger scores before translating the coefficient sign into environmental language.
Glossary

Key terms

Covariance
Covariance measures whether paired variables tend to deviate from their means in the same or opposite directions.
Pearson Correlation
Pearson Correlation is a standardised measure of linear association between paired quantitative variables.
Monotonic Association
Monotonic Association means one variable tends to move consistently upward or downward as the other increases.
FAQ

Correlation and Ranked Relationships FAQ

Can Pearson's r miss a relationship?

Yes. Pearson's r targets linear association, so a strong curved pattern can produce a coefficient near zero when opposite slopes cancel. Inspect the scatterplot and use a scientifically justified nonlinear model when the geometry demands it.

When is Spearman's rho useful?

Spearman's rho is useful for ordinal variables or a monotonic pattern when equal quantitative gaps and linearity are unsuitable. It works with ranks, so interpret ordered association rather than a fixed response change for one physical unit.

Does a large correlation show causation?

No. A coefficient describes co-variation in the paired sample. Confounding, reverse direction, common trends, selection and measurement can create the pattern. Causal interpretation needs design and evidence that distinguish those competing explanations.

Study strategy

Assessment move

Sketch positive, negative, near-zero and curved point clouds, then predict Pearson and ranked coefficients. For each output, write one statistical sentence and one separate environmental sentence that names confounding or design limits without denying the observed association.

Working through Correlation and Ranked Relationships in 7SSGN110? Sia is AskSia’s AI Statistics tutor — ask any 7SSGN110 Correlation and Ranked Relationships question and get a clear, step-by-step explanation grounded in how 7SSGN110 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 18 of your subjects at King's College London, plus 1,000+ more Bibles across the AskSia library.
Sia - your 7SSGN110 tutor, unlimited, worked the way the exam marks it
The full 4-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works