StatReasonWhat this answers
Spearman's rho answers "as one variable increases, does the other tend to consistently increase (or decrease) as well, regardless of whether that relationship is a straight line?" Unlike Pearson correlation, which measures only linear association, Spearman detects any consistent monotonic pattern, including curved relationships where a value always increases with the other, just not at a constant rate.
When to use it, and when not to
Use Spearman when your data is ordinal, contains extreme values that would distort a Pearson correlation, or shows a clearly curved but still consistently increasing (or decreasing) pattern. If your relationship is genuinely linear with no extreme values, Pearson correlation uses more information and is the more standard choice.
How it is calculated
Every x value is converted to its rank among all x values, and every y value to its rank among all y values (tied values receive the average of their tied rank positions), then Pearson correlation is computed on those ranks rather than the raw values. A significance test uses a t distribution with n minus 2 degrees of freedom. With fewer than about 500 pairs, this calculator flags the result as using an asymptotic approximation, since an exact permutation-based p-value would be more precise at very small sample sizes.
Worked example
For x equal to 1 through 5 and y equal to their squares (1, 4, 9, 16, 25): the relationship is clearly curved, not a straight line, so Pearson's r would be slightly below 1. But because y strictly increases every single time x increases, the ranks of x and y match perfectly, giving a Spearman's rho of exactly 1, correctly capturing the perfect monotonic pattern that Pearson alone would understate.
Assumption audit
What this result does not mean
Correlation, Spearman or Pearson, does not establish that changes in one variable cause changes in the other. A significant p-value means the data provide evidence that rho differs from 0, not that the monotonic relationship is strong; check the actual value of rho for that.
Limitations
Spearman only detects monotonic relationships. A relationship that rises then falls (or vice versa) can produce a Spearman's rho near 0 even though a real, strong pattern exists, since ranks cannot capture a reversal in direction.