Correlation for ranks and curves
Spearman's rho answers the same question as Pearson's r, with one change of input: each value is replaced by its rank. First becomes 1, second 2, and so on, with tied values sharing the average of the ranks they span. Pearson's formula is then applied to the rank pairs. The result measures monotonic association: whether the second variable tends to rise when the first rises, at any pace, in any curve shape, as long as it keeps rising.
Two judges rank eight gymnasts. Judge A: 1 2 3 4 5 6 7 8. Judge B: 2 1 4 3 6 8 5 7. The rank pairs give ρ = 0.833: strong agreement, with the swap of adjacent places doing little damage. A t-approximation gives p ≈ 0.010, though for eight pairs an exact table is the safer reference.
Ties, handled properly
Ties are not a special case bolted on; the average-rank rule absorbs them. If two products tie for 5th and 6th, both receive rank 5.5. Six contestants ranked 1, 2, 3, 4, 5.5, 5.5 by one judge against a clean 1 to 6 from another give ρ = 0.928. The calculator applies this automatically. Plenty of short formulas for Spearman ignore ties and quietly go wrong when they appear; the rank-then-Pearson method used here stays exact.
Using Spearman as a weaker Pearson. It answers a different question. Feed it a straight-line relationship and it returns a similar value; feed it a steepening curve and it returns near 1 where Pearson sags. Neither is the correction of the other.
When ranks are the right data
Judges' scores, league positions, Likert-scale survey answers, and any data where the gaps between values carry no meaning: these are Spearman's home ground. It is also the standard rescue when one wild outlier would otherwise run the analysis: the outlier becomes just the top rank, no bigger than any other rank step.
Frequently asked questions
When should I use Spearman instead of Pearson?
When your data are ranks or ordered categories, when the relationship is steady but curved, or when outliers would dominate Pearson's calculation. If the relationship is linear and the data are clean measurements, Pearson is more informative.
How are ties handled?
Tied values each get the average of the ranks they occupy. Two values tied across ranks 5 and 6 both become 5.5, and the total of the ranks is preserved.
Can Spearman's rho be exactly 1 without a straight line?
Yes. Any perfectly increasing relationship, however curved, gives rho = 1, because the rank order is perfectly preserved. That is the point of the method.
Is the p-value exact?
The calculator uses the standard t approximation. For very small samples (under about 10 pairs), exact permutation tables are preferable and the approximation should be read as a guide.