UPCOMING WEBINAR: The complete guide to analyzing NPS When categories are merged by dragging and dropping on a table, the new categories do not appear when the variable is selected to create a weight. For example, if you have age collected as a Numeric variable, but you have turned it into a Nominal variable and merged all its categories, the original pre-merged categories will appear when you select this variable as an Adjustment variable in a Weight. This is to ensure that if people merge categories or make overlapping categories, the weights in their project are not inadvertently changed.
Please note these steps require a Displayr license.
Method
There are a number of ways of addressing this:
Method 1 - Drag and drop categories to group them in the weighting dialog
Instead of merging the categories in the table of targets, we can do this in weighting dialog itself.
1. Bring up the weighting module as usual via Anything > Weight or by hovering over any variable in your Data Sources tree > Plus (+) > Weight.
2. Select your adjustment variable.
3. Select the category label in the weighting table that you want to combine and drag it onto the appropriate category label. Here, for example, I have dragged 25 to 34 onto 18 to 24 to effectively create 18 to 34:
Method 2 - Create an R variable that keeps the changes made in the table of that variable
1. Combine the categories in the table of the variable in question, e.g., 18 to 24 and 25 to 34 as 18 to 34:
2. Select the modified variable in the Data Sources tree.
3. Click the Plus (+) that appears when you hover over it.
4. Select Custom Code > R - Numeric.
5. Either select or drag the variable from the Data Sources tree to the R CODE field in order to populate the variable name. Otherwise directly enter or paste the variable name there.
6. Change Structure to Nominal.
7. Update Name and Label.
8. Create your weight as usual, but select the new variable and it will retain the modified categories:
Method 3 - Use a method for merging the categories which change the underlying values
See How To Band Numeric Variables Using JavaScript or How to Band Numeric Variables Using R).