Show TOC
## Outlier Correction with the Median
Method

### Use

### Features

### Activities

The median method is one of two methods for outlier correction.

The system uses the median method to determine the ex-post forecast values for the basic value, trend value, and the seasonal index. It can thus calculate an expected value for each historical period

The tolerance lane
is calculated as **Sigma * Expected value**.

Note that the median method does not use measures of error. This means that the tolerance lane values are absolute and do not depend on the scatter of the data.

There are no Customizing settings for this method.

In the univariate forecast profile or directly in univariate forecasting (on the Settings tab page) select Median Method in the Outlier Correction field.

When you execute a forecast, the system:

...

1. Carries out a forecast with the chosen model using the ‘raw’ historical data.

2. Determines the basic value, trend value, and seasonal indices using the median forecasting method as described in the Activities section of Median Method.

3. Determines an expected value for each historical period, as described above.

4. Calculates the tolerance lane for each historical value.

As the tolerance lane is determined directly from the expected value and not, as in the ex-post method, from an error measure (MAD), using the same value of sigma as in the ex-post forecast may produce unexpected results. In general you should enter a smaller value for the median method than for the ex-post method.