This image shows the formula for the Mean Error, also known as ME. First the sum is taken, for values from 1 to n, of the exact, sign-significant differences between each actual likelihood and the corresponding predicted likelihood. Then the sum is divided by n.