What Is Correlation Analysis? Interpreting r, Pearson vs. Spearman
Daniel Ha · Seoul National University PhD StudentWhat Is Correlation Analysis?
Correlation analysis summarizes how much two continuous variables vary together — the direction and strength of their linear relationship — in a single coefficient, r. Values near +1 indicate a strong positive relationship, values near -1 a strong negative relationship, and values near 0 a weak linear relationship.
When Should You Use Correlation Analysis?
Use it to establish whether and in which direction variables are associated, before making any causal argument. It is also routinely used to explore relationships ahead of a regression analysis and to screen for multicollinearity.
How Do You Interpret a Correlation Coefficient?
The sign gives the direction and the absolute value gives the strength. Thresholds vary by field, but the following is a common reference.
| |r| range | Strength |
|---|---|
| 0.0 to 0.2 | Very weak |
| 0.2 to 0.4 | Weak |
| 0.4 to 0.6 | Moderate |
| 0.6 to 0.8 | Strong |
| 0.8 to 1.0 | Very strong |
Check the p-value to confirm the relationship is statistically significant, and report r² for the proportion of variance explained.
What Is the Difference Between Pearson and Spearman?
Use the Pearson correlation when both variables are continuous and meet normality and linearity assumptions, and the Spearman correlation when the data are ordinal or fail normality, since it works on ranks. Spearman is also more stable when outliers exert heavy influence.
What Should You Be Careful About in Correlation Analysis?
Correlation is not causation. A significant correlation may be driven by a third variable, so describe it as an association rather than an effect. Outliers and nonlinear relationships also distort r, so always inspect the scatterplot alongside the coefficient.
Summary
Correlation analysis summarizes the direction and strength of a linear relationship between two variables as r. Choose Pearson or Spearman according to scale and normality, and interpret carefully as association rather than causation. When you need effect size and prediction, move on to regression analysis, and for causal structures among several latent variables, move to structural equation modeling.