Independent Variables for Multiple Regression

Multiple Regression uses regressors (other data sets) to create a model and define the relationship between the dependent variable (the item's history) and the specified regressors (independent variables) for the ATT forecast engine. You can specify these regressors on the Independent Variables tab of the Forecast Engines page.

See Defining ATT forecast engine.

The tab contains a list of measures that include the independent variables mapped to the forecast engine, for a multiple regression analysis. The measures that can be used to store the independent variables results are also listed.

This table shows the applicable measures:

Measure Description
Independent Variable The measure is used to store the independent variable that must be transferred to the forecast engine. Variables contain the data for the cycle period horizon (history and future horizons).
Regression Coefficient The measure is used to store the regression (or beta) coefficient data that is generated for the corresponding independent variable. This represents the percentage of the corresponding independent variable that is built in the model. This is a static (time independent) value that is written to the PCONST constant.
t Statistic The measure is used to store the 't Statistic' data that is generated for the corresponding independent variable. The measure represents the coefficient divided by the standard error. These are used to test each individual independent variable for relevance. If the generated data is large (absolute value > 2) then the independent variable is statistically significant when compared to zero at the 95% level, and must be accepted. This is a static (time independent) value that is written to the PCONST constant.
t Statistic P-Value The measure is used to store the 't statistic P-Value' data that is generated for the corresponding independent variable. The P value is the probability of exceeding the observed 't statistic', if the true coefficient is zero. A low value (< 0.05) implies statistical significance. This is a static (time independent) value that is written to the PCONST constant.
Variance Inflation Factors The measure is used to store the variance inflation factor data for the corresponding independent variable. These are used to measure multicollinearity, that is two or more variables in the model are highly correlated. This occurs when independent variables are linearly related to each other and the dependent variable.