A Regression Target gives you the capability to measure the potential effect a change on your operation will have. Additionally, they give you the ability to compare two or more variables. Regression targets work by comparing two parameters and determining the correlation between the variables. These variables may be consumption vs. production, consumption vs. cost, or pressure vs. production. The correlation will be used to create an equation which can be used to create a dynamic target.
Example: Electricity consumption at a facility is a variable of interest that you would like to track and find savings for. You find a relatively high correlation (> 0.75) between electricity consumption and production at the facility (Variable). You start an initiative at their facility to try to reduce electricity consumption. However, production at the facility is reduced for the following 2 months due to another independent event at the facility. Having previously created a regression target using the correlation between production and electricity consumption, the regression model will still capture any energy savings of the electricity consumption despite the production variable changing.
To create a new regression target, you can select the Regression icon. You will be directed to a page where you can fill in the variables that you wish to use as comparison in the regression model.
Parameters can be added by clicking the Add button underneath the Input Parameters panel.
A list of all available variables will appear. You can search keywords in the search bar, or manually scroll through the list of available variables. Clicking a variable will bring the variable into the regression analysis. Several variables can be added to a regression target to form a multi-variable regression target.
Several statistics are created once the parameters are added. These statistics primarily advise on how high the correlation between the variable and inputted parameter/variable is.
A primary indicator of correlation is the R2 value generated. A regression analysis that outputs an R2 of 1.00 is a perfect correlation. A correlation between two variables that generate an R2 closer to 1.00 indicates the regression model fits the data well.
