Forecast Errors And Summary Statistics
The forecast error is equal to the actual minus forecast.
Some Frequently Used Summary Statistics Include:
Mean Error =
Mean Error can indicate that errors are "centered" because positive and negative errors cancel each other out. For example, if you have one error of +10,000 and another of -10,000, then the mean error is zero.
Mean Absolute Error (MAE) =
Taking the absolute value of each error prevents positive and negative errors from canceling each other out. For example, if you have one error of +10,000 and another of -10,000, then the mean absolute error is 10,000.
Mean Squared Error (MSE) =
Taking the square of each error prevents positive and negative errors from canceling each other out, and it also amplifies the impact of “big” errors.
Mean Absolute Percentage Error (MAPE) =
Expressing the error as a percentage can make it easier to judge the error's impact. You could also calculate Mean Percentage Error or Mean Squared Percentage Error.
Selecting A Summary Statistic:
Each summary statistic provides information about the nature of the forecast errors. It's a good idea to check a few of these summary statistics and compare them for each model you test. However, you'll probably want to select one to report to your boss -- too many numbers can be confusing and only increases the liklihood of awkward questions arising. So, the choice of which summary statistic to report can boil down to:- how easy it is to communicate to senior managers
- volatility in relation to call volume
site navigation
Planning Tool Overview >> Forecasting Overview <<
Forecasting Process Forecasting Models >> Error Statistics << Example
Neville Shea * R.D. 1 Box 263, Falls PA 18615 * (570) 333-4913 * nevshea@earthlink.net