Grouped Characteristic

 

Grouping is used for data that is more explanatory in bins than by the individual data values. For example, vehicle odometer readings or year built data are typically grouped into bins.

 

If the Grouped data type is chosen for a characteristic, Cognalysis MultiRate will automatically group the data. The default is 10 groups (including a Null group, even if there are no NULL values) and the bins will have as close to equal exposure amounts as is possible:

 

 

To view and/or edit the groupings via the window pictured above, click on the gray "..." button to the right of the desired grouped characteristic in the Data and Analysis Control panel:

 

 

The predetermined groupings can be overridden in two ways:

a.Adjust the number of Groups

To create fewer bins or more bins, type in the desired number of groups in the “New Group Count” box and click “Redetermine Groupings” button. The software will automatically determine the new groups, again attempting to keep the exposure in each bin relatively constant. Note that a NULL bin is always required and counts as one of the bins regardless of whether there are NULL values in the data.

b.Enter Bin Limits Manually

Breakpoints can also be entered manually for any of the bins. The value listed in the "Characteristic Value Upper Limit" column is the upper bound for the bin. Click on the value so that a black box appears, then type in the desired upper breakpoint and hit enter. The system will automatically calculate and update the new amount of exposure and target in each bin. Note that the values must be strictly increasing.

 

          Ex) Original first few bins:

                    

 

                     After adjusting the upper limit of the 2nd bin to 100:

                    

 

Grouped fields are typically numeric, but do not have to be. For example, if there is a field whose value is any of the letters A through Z and neighboring values (C & D, for example) are known to have a stronger relationship than values farther away from each other (C & Z, for example), then grouping the characteristic may make sense, especially if fewer than 26 bins is desired. Also keep in mind that not all numeric fields make sense to be grouped.