Understanding Mean Absolute Deviation (MAD) in Forecasting

In forecasting, evaluating the accuracy of predictions is crucial. Various metrics help quantify the difference between forecasted values and actual observed values. The Mean Absolute Deviation (MAD) is one such metric.

What is Mean Absolute Deviation?

The Mean Absolute Deviation (MAD) measures the average magnitude of the errors in a set of forecasts, without considering their direction. It calculates the average of the absolute differences between the actual values and the predicted values over a given period.

Calculating Mean Absolute Deviation

The formula for MAD is as follows:

$$ \text{MAD} = \frac{\sum_{i=1}^{n} |A_i - F_i|}{n} $$

Where:

• $A_i$ represents the actual value for period $i$.
• $F_i$ represents the forecasted value for period $i$.
• $|A_i - F_i|$ represents the absolute error (the magnitude of the error) for period $i$.
• $n$ is the total number of periods.

By taking the absolute value of each error ($|A_i - F_i|$), MAD ensures that both positive errors (forecast higher than actual) and negative errors (forecast lower than actual) contribute positively to the total deviation. This means MAD focuses solely on the size ormagnitude of the error, not on whether the error was an over-forecast or an under-forecast.

Analyzing the Options

Let's examine how MAD relates to the provided options:

• Gross error: While MAD represents error, it's an average and smoothed measure, not typically referred to as "gross error," which might imply outlier or very large errors.
• Direction and magnitude of the error: MAD explicitly ignores the direction of the error due to the absolute value function. Metrics like Mean Error (ME) consider direction.
• Direction of the error: MAD does not express the direction. The sum of signed errors, divided by the count (Mean Error), captures direction.
• Magnitude of the error: This aligns perfectly with the calculation of MAD. The absolute value function ensures that only the size (magnitude) of the difference between actual and forecast is considered and averaged.

Conclusion on MAD

Therefore, the Mean Absolute Deviation specifically expresses themagnitude of the error in forecasting.