When to Use Spearman vs Pearson
You'll learn the key decision criteria for choosing between Spearman's and Pearson's correlation coefficients.
The Core Decision
Both Spearman and Pearson measure relationships between variables, but they apply to different situations. The choice depends on four key factors: your data type, distribution shape, presence of outliers, and the nature of the relationship you're measuring.
Decision Criteria
1. Ordinal Data
If your data is ordinal (ranked categories like "low," "medium," "high"), use Spearman. Pearson requires interval or ratio data with meaningful numerical differences.
2. Non-Normal Distributions
When your data is heavily skewed or doesn't follow a normal distribution, Spearman is more robust. Pearson assumes roughly linear patterns that work best with normally distributed variables.
3. Outliers
Outliers can dramatically inflate or deflate Pearson's r because it's sensitive to extreme values. Spearman, working with ranks instead of raw values, resists outlier influence—think of it like converting "100,000" both to ranks where the gap matters less.
4. Monotonic vs Linear Relationships
- Pearson detects linear relationships: as X increases by a constant amount, Y changes by a constant amount
- Spearman detects monotonic relationships: as X increases, Y consistently increases (or decreases), but not necessarily at a constant rate
Imagine temperature and ice cream sales (linear) versus wealth rank and happiness (monotonic but not perfectly linear)—Spearman catches the second pattern better.
Quick Decision Guide
Use Pearson when you have: continuous data, normal distributions, no major outliers, and expect a linear relationship.
Use Spearman when you have: ordinal data, non-normal distributions, outliers, or monotonic (but not necessarily linear) relationships.
Key Takeaway: Choose Spearman for ranked, skewed, or outlier-prone data detecting monotonic relationships; choose Pearson for normally distributed continuous data detecting linear relationships.