Correlation vs Causation: Pearson's r and Spearman's Rho
Linear versus monotonic associations, confounding variables, Anscombe's quartet, and outlier sensitivity.
In this lesson you will learn
- 1Explain why a strong correlation coefficient (r or ρ) cannot establish causation in the presence of unmeasured confounders.
- 2Choose between Pearson's linear r and Spearman's rank-based ρ based on normality, linearity, and outliers.
- 3Recognise why r = 0.00 rules out only a linear relationship, not a U-shaped curvilinear relationship.
On this page▼
Linear vs Monotonic Association & Confounding
This section contrasts Pearson's r against Spearman's ρ and illustrates how scatterplots expose non-linearity, outliers, and spurious confounding.
Reporting Pearson's r without inspecting a scatterplot can hide a strong curvilinear relationship or let a single outlier create a false correlation.
A correlation coefficient summarises the direction (− or +) and strength (from 0.0 to 1.0) of association between two variables:
| Metric / Concept | Definition | Clinical Guidance |
|---|---|---|
| Pearson's r (−1 to +1) | Standardised covariance measuring how tightly points cluster along a straight line. | |r| ≈ 0.10 weak, 0.30 moderate, ≥ 0.50 strong linear association. |
| Spearman's rho (ρ) | Pearson's correlation computed on the ranks of X and Y (monotonic trend). | Use when either variable is skewed, ordinal, or contains outliers. |
| Confounding Variable | An extraneous factor associated with both exposure X and outcome Y that creates an apparent correlation. | Control via randomization or multivariable regression adjustment. |
Interactive Correlation Lab: Pearson's r vs Spearman's ρ
Compare linear Pearson r against rank-based Spearman ρ under outliers and curved patterns
Clinical Example
A hepatology study correlates cumulative drug dose with serum ALT across 40 patients, including two acute hepatic injury outliers with ALT > 600 U/L.
Interpretation: Because enzyme elevation accelerates non-linearly at high doses, Pearson's straight-line fit underestimates the relationship (r = 0.31), whereas Spearman's rank correlation accurately captures the monotonic rise (ρ = 0.68).
Common Pitfall
Concluding that two variables are completely unrelated when Pearson's r ≈ 0.00, without checking whether a U-shaped curve is present (such as hypothermia and hyperthermia both increasing mortality).
Included with Any Analysez AI Study Package
This deep interactive chapter (Correlation vs Causation: Pearson's r and Spearman's Rho) and its clinical simulation labs are unlocked for everyone who has bought any Analysez AI study package.