Creating a Time Series Variable

Use

As variables in an MLR model, you can use either key figures from the planning area or time series. This procedure describes how to create a time series variable.

Prerequisites

You have made all the required entries on the *Maintain Forecast Profile* screen where the master forecast profile is defined.

You have made all the required entries in the group boxes *Model for multiple linear regression* and *History for dependent key figure Y*.

Procedure

- On the
*MLR Profile*definition screen, click on*Time series*(*Create or change time series*)*Selection*group box at the bottom of the screen. - Choose
*Enter*. - Enter a time series ID and a short description of the time series variable.
- Enter the date on which the time series variable comes into effect; that is, the date to which the first period of the variable should be pegged.
- Enter the number of decimal places in which the time series value should be saved and displayed.
- Enter the number of periods.
**You must enter****at least as many periods****as the combined number of periods in the future and historical time horizons**(as defined in the master forecast profile). - Select
*Edit time series*. - Enter the variable's values.

This date is a default. You can overwrite the default with another date in the *MLR Profile* definition screen.

If you do not set a start date, the system uses as the start date the beginning of the future planning horizon from the master forecast profile.

You can use the same time series variable in several forecast profiles with time horizons of different lengths. In one profile, the time series variable may have the same number of periods as the combined number of periods in the time horizons. In a second profile, the time horizons may be shorter; therefore, the time series variable contains more periods than are needed for MLR. When the system carries out causal analysis based on this second profile, it ignores the periods in the time series variable that go beyond the end of the future time horizon.

The system uses the historical values to build the MLR model. It uses the future values to run the forecast.

- To quickly replicate patterns, copy an interval of values already entered to an interval with no values by choosing
*Copy interval*. - To duplicate all periods of the time series, choose
*Update time series*.

- Choose
*F8*.