The mean absolute error measures how far estimates or forecasts differ from actual values. It is most often used in a time series, but it can be applied to any sort of statistical estimate. In fact, it could be applied to any two groups of numbers, where one set is “actual” and the other is an estimate, forecast or prediction. Alternatives include mean squared error, mean absolute deviations and median absolute deviations.

Set up your data in two columns. One column should have the predicted values, or estimated values, the other the actual values.

Subtract the predicted value from the actual value in each row.

Take the absolute value of each difference you calculated in Step 2. So if the difference is negative, remove the negative sign. If it is positive, leave it as is.

Add up the absolute values.

Divide by n -- that is, the total number of rows.