Spearman

Definition

Spearman's Rank Correlation Coefficient is a non-parametric measure of rank correlation. It assesses how well the relationship between two numerical features can be described using a monotonic function, which is a relationship that does only one of the following: increase or decrease.

Range of score: -1 to 1

A coefficient of 1 indicates a perfect increasing monotonic relationship, whereas one feature increases, the other does as well. A coefficient of -1 indicates a perfect decreasing monotonic relationship, whereas one feature increases, the other decreases. In a perfect increasing/decreasing monotonic relationship, “perfect” relates to the consistency of direction of change (and not the magnitude). A coefficient of 0 indicates no monotonic relationship, meaning that there's no consistent trend in the ranks of the features.

How it works

Spearman's correlation is calculated by converting each feature to ranks (i.e., the first, second, third, etc. highest value), then computing the Pearson correlation coefficient on these ranks. This makes it less sensitive to outliers and robust to non-linear relationships, as it only considers the order of the values, not their actual magnitudes.