Purpose

In this process you specify the parameters that the system uses for causal analysis with multiple linear regression (MLR). These are:

- The key figure for the historical values of dependent variable (Y)
- The independent variables (X
_{i}) - The method used to measure the error in the historical values
- Assumptions on how the error is distributed
- A diagnosis group to specify the tolerances for the measures-of-fit.
- Time shifts between dependent and independent variables.

Prerequisites

You have created a master forecast profile.

Process Flow

- You enter a name and description for the MLR profile
- You specify the key figure for the historical data
- You enter information on where the system can find the data for the independent variables. This can be from either:

- Any key figure in the planning area

If you use key figures, you can define a time shift between cause and effect. For example the demand for a consumer product may depend on the income tax rate. However since most employee are paid at the end of the month, it takes at least one month till the effect of a reduction in the tax rate will be seen.

- A time series. For details on how to create a time series for MLR, see Creating a Time Series Variable.

Diagnosis Group

If you want the system to generate alerts when measures of fit are not reached or exceeded, enter a diagnosis group. See also Measures of Fit.

Measured Value Error

In certain circumstances it is useful to be able to determine how the system determines the error in a measured value. The system can assume a normal or a Poison distribution. In the case of a normal distribution, you can specify whether the error variance is constant or variable. If it is constant, you can enter a value in the *Sigma* field. If it is variable, the system calculates the necessary values.

In general you do not need to make an entry in this field. For more details, refer to the F1 help.

.