MLR Forecast with weighted least square method

[israel_cano]

Hi, We are implementing MLR Forecasting, since the kind of industry and
process, is poultry, we are testing different variables. In our tests
outside APO, we reach better results with the method of weighted least
square. APO uses Ordinary least square.

We know that we can use any model with the help of a Badi, but since in
the SCM 7.02 the engine for MLR is ABAP based. ¿Could you give us some
recomendation of the best way to implement weighted least square if we try to use this method ?

Regards

[former_member351430]

Hello Israel,

One way is to use macros to transform the input data to the MLR regression model. Specifically, you should do the following:

1. Multiply the dependent variables by the square root of weight of each observation

2. Multiply the dependent variables (factors) by the square root of the weight of each observation

3. Include a new dependent variable (factor) with values equal to the weight of each observation

4. Force the intercept to be zero when running the multiple linear regression model