How to Analyze the Number of Selections in a Multi-Select Question

Sometimes, with a multiple-response question (i.e., a Binary-Multi variable set), you may want to show the average number of options selected (or items mentioned) across respondents. In order to calculate this, you must first create a new variable that is the Count of the number of selections per respondent. Then, when you use this Numeric variable in a built-in drag-and-drop table, Displayr automatically shows the Average (or any other standard statistic). Similarly you may have a grid question that shows attributes selected per brand and you want to see the number of attributes associated with each brand, see Additional Notes below for guidance for this.

The example below is a multi-select question; if respondents are asked which of the following list of colas they are aware of, it may be of interest to count the number of colas each respondent selected.

Which of the following colas are you aware of?
_Coca-Cola
_Diet Coke
_Coke Zero
_Pepsi
_Diet Pepsi
_Pepsi Max

Requirements

• The multi-response variable set must be structured as Binary-Multi, see How to Create New Binary Variable(s) for guidance.

Method

1. In the Data Sources tree, select a variable set or group of variables.
2. Hover and click + > Calculate Across Variables > Count. This will create a new variable that is the Count of the 1s (coded for Selected) across the variables. Missing data is ignored by default.
3. Drag the new variable to a Page or the Report tree to make a table, and you'll see the Average number of selections.
   
4. OPTIONAL: You can also see how many respondents chose a certain number of options. Select the count variable (or a Duplicate of the variable) and in the object inspector , change the Structure > Nominal: Mutually exclusive categories. You'll see the proportion who chose each number. 

Additional Notes

If your data is messy or in a Nominal-Multi structure, you can customize which values to count in the object inspector > Data > Values to count field of the count variable if needed.

Binary-Grid

If your data is setup in a Binary-Grid format, you can also count selections for each row or column. For a variable set where the Brands are in the Columns and Attributes are in the Rows, you can select variables for one brand at a time and use the + > Calculate Across Variables > Count automation to calculate selections for each brand. You can then Combine those together into a Numeric-Multi. 

Depending on how clean your labeling is, you may be able to use a shortcut. Because Calculate Across Variables functions also match labels between variable sets, a quicker way to do this for all brands is to select the Binary-Grid and use + > Convert To > Binary-Multi > Split Grid by Rows (if brands are in columns). You'll have a variable set for each attribute with variables for each brand. Then with all of those selected, click + > Calculate Across Variables > Count. This will match the brand variable in each attribute set and add up how many sets/attributes it was selected on to make the Numeric-Multi set.

Next

How to Create New Binary Variable(s)