Use this activity to create, remove, select and view a graphical and table presentation of explanation variables. To select which explanation variable that is displayed in the graph you must move the table cursor to a cell in the explanation variables column. Then that explanation variable is shown in the graphs. There are 2 graph tabs, the Explanation Variable Graph tab and the Correlation Graph tab. The explanation variables are used with the Multiple Regression Forecast Model.
When adding new explanation variable it is possible to base the variable on an existing forecast part, this parts adjusted demand and forecast will then determine the explanation variables numbers in the historical range and forecast range. You can also set the access level of the explanation variable for example make the explanation variable to visible on the current flow only or only visible on the forecast part/flow. Default all explanation variables are global.
The x-axis unit is time (periods) from the first period of historical data for the current item to the last period of the forecasting horizon. The choice of period unit (week, month or quarter) is done once and for all in the Period Version Tab in the Demand Plan Server setup. The left y-axis is the historical data and forecast for the selected forecast part. The right y-axis is for selected explanation variable. The y-axis is automatically scaled to fit the data that is presented. If the window is too small to fit all the data into the x-axis, then scrolling buttons are activated in the bottom left and right corners of the window.
Graph Adjustments
The Data in Explanation variable graph can be changed directly in the graph by dragging the graph elements. The system displays the on-line value of the periodical elements as they are dragged if the Notes function is activated.
| Key | Name | Description |
| - | Single period | Drag an individual forecast element up or down directly in the graph to raise or lower the forecast value, while holding down the left mouse button. |
| Shift | Straight line | Drag an index value to the right while holding the Shift key to arrange the elements in a straight line from the initial position of the period to where it was finally dropped. |
| Ctrl | Level | Hold the Ctrl key while dragging any forecast element up or down to raise or lower all elements of this type equally. |
| Shift+Ctrl | Free Draw | Hold the Shift+Ctrl key while dragging any forecast element up or down to draw a freeform line of forecasts. |
This is a Scatter Graph. The x-axis unit is historical demand from the current selected part or group, the unit is the standard unit for the Part No (number). The left y-axis is the selected explanation variables unit. Each ball represents a adjusted demand / explanation variable data point. The black thin line is the regression line trough the scatter points, a positive slope indicates positive correlation between the part/group and the explanation variable, while a negative correlation will give a negative slope of the regression line. The distance the points have from the correlation line indicates how strong the correlation is, the closer the ball's are to the correlation line the stronger (high correlation factor) the correlation.
This is a table showing the forecast part's adjusted and system forecast and the adjusted demand and all the explanation variables connected to the selected forecast part. The explanation variables can be changed in this table as well as in the explanation variable graph.
| Rows: | The top row called included, it is in this row you can add/remove explanation variables that is to e included in the multiple regression model for this part/group. In addition to this one top row there is a row for each periodic value of the data elements. The time interval extends from the first period having existing data on the current part/group to the last period of the forecast horizon. Labels are in the following format: YYYY-PP (where Y = year and P = period). |
| Columns: | One column for each existing explanation variable, and one row with the adjusted demand, system forecast and adjusted forecast |
This table shows the correlation factor between the selected forecast part's adjusted demand and the explanation variables and between the different explanation variables. The matrix is a square matrix and the diagonal of this matrix will be all 1.0 since a vector's correlation factor with itself is 1.0. Each cell in this matrix shows the correlation between the corresponding X and Y vectors. A correlation factor is a number between -1 and 1 that tells the strength and direction of a relationship between to variables.
The correlation matrix, this matrix shows the correlation between the different explanation variables and the other explanation variables and the adjusted demand of the selected forecast part/group. Here the Row labels indicates the explanation variable's the column values in this row is correlations coefficient for the explanation variable with the other explanation variables. A rule of thumb any correlation coefficient value (cell number) above 0.7 indicates multicollinearity.
In some sense, the collinear variables contain the same information about the dependent variable. If nominally "different" measures actually quantify the same phenomenon then they are redundant. Alternatively, if the variables are accorded different names and perhaps employ different numeric measurement scales but are highly correlated with each other, then they suffer from redundancy. For example the 2 explanation variables number people listening to your radio advertisement and the number of dollars spent on advertising has a correlation coefficient on 0.85 these to variables explains the same data and only one of them is needed.
Consequences of multicollinearity
In some cases it will not be possible for Demand Planning to calculate the regression. In the presence of multicollinearity, the estimate of one explanation variable's impact on y (the forecast) while controlling for the other explanation variables tends to be less precise than if the explanation variables were uncorrelated with one another. The usual interpretation of a regression coefficient is that it provides an estimate of the effect of a one unit change in an explanation variable, X1, holding the other explanation variables constant. If X1 is highly correlated with another explanation variable, X2, in the given data set, then we only have observations for which X1 and X2 have a particular relationship (either positive or negative). We don't have observations for which X1 changes independently of X2, so we have an imprecise estimate of the effect of independent changes in X1.
A explanation variable has been deleted, changed or a new explanation variable has been created.