| Model Browser User's Guide | |
The Define Test Groupings dialog collects records of the current data set into groups; these groups are referred to as tests. Test groupings are used to define hierarchical structure in the data for two-stage modeling.
You access the dialog from the Data Editor by doing one of the following:
When you enter the dialog, no plot is displayed.
) adds the currently selected variable in the Variables list to the list view on the left. Alternatively, double-click variables to add them.
You can now use this variable to define groups in the data.
In the following example, the variable n is being used to define groups. The maximum and minimum values of n are displayed.
The Tolerance is used to define groups: on reading through the data, when the value of n changes by more than the tolerance, a new group is defined. You can change the Tolerance by typing directly in the edit box.
You can define additional groups by selecting another variable and choosing a tolerance. Data records are then grouped by n or by this additional variable changing outside their tolerances.
You can plot variables without using them to define groups by clearing the Group By check box.
You can remove variables from consideration by selecting the unwanted variable in the list view (the selection is highlighted in blue) and clicking the Remove variable button 
.
The plot shows the scaled values of all variables in the list view (the color of the Tolerance text corresponds to the color of data points in the plot). Vertical pink bars show the tests (groups). You can zoom the plot by Shift-click-dragging or middle-click-dragging the mouse.
Test number variable contains a drop-down menu showing all the variables in the current data set. You can select any of these to number the tests (for example, lognumber could be useful (instead of 1,2,3...) if the data was taken in numbered tests and you want access to that information during modeling).
Every record in a test must share the same test number to identify it, so when you are using a variable to number tests, the value of that variable is taken in the first record in each test.
Test numbers must be unique, so if any values in the chosen variable are the same, they are assigned new test numbers for the purposes of modeling. (This does not change the underlying data, which retains the correct lognumber or other variable.)
Reorder records allows you to reorder records in the data set. This sorts records before grouping. Otherwise, the groups are defined using the order of records in the original data set.
Show original displays the original test groupings if any were defined.
One test/record defines one test per record, regardless of any other grouping. This is required if the data is to be used in creating one-stage models.
Clicking OK accepts the test groupings defined and dismisses the dialog.
| | Storage | Data Wizard | |
