Terms from Statistics for HCI: Making Sense of Quantitative Data

The covariance is the average product of two variables after their means have been subtracted: cov(x,y) = ( Σ (xi - μx) × (yi - μy) ) / N. It is a generalisation of the variance of a single variable. It measures how large is the common variability of the two variables; if they are unrelated it is near zero. For example, if you measured the heights and weights of people you would expect the covariance to be positive. The size of the covariance is also influenced by the level of variability of the two individual variables: for example, the covariance between length and weight of sparrows would be a lot smaller than the same calculation for whales. The Pearson correlation coefficient corrects for this by dividing by the product of the standard deviation (s.d., σ)¬†of the two variables.