Categorizations

Categorizations are variables that are composed of several questions that have the same scale. The categorization can present the results from these questions together with an average calculated from the questions it contains (go to Setting up Categorizations for more information).

Working with the categorization object is very similar to working with a grid question in the Table Designer, but the categorization has one special property: the “Calculation Rule” . The calculation rule property defines how the average of the elements is to be calculated.

Figure 1 - The Properties sheet for a Categorization

When a categorization is added to a table, two methods of calculating the average are available:

• Avg. of Aggregates (the average of the aggregated averages) - first the averages of all the individual elements are calculated, then a simple average of these results is calculated by adding all the element’s averages together and dividing by the number of elements.
• Avg. of Individuals (the average of individuals) - the averages are first calculated for each respondent, then an overall average is calculated by calculating an aggregated average of the averages for all respondents.

These two methods will return equal results if all respondents have answered all questions. However if not all of the questions are required, or elements in the answer list have not been assigned weights (for example “Don't know”), then there may be differences between the results for the two methods of calculating the total average. In the example belowhere, the average is calculated on three questions, q1, q2 and q3, using the two different methods. The results differ because some respondents have not provided answers for all the questions.

Figure 2 - Example of different average calculations for categorizations

Below Here is an example of a categorization set up based on a grid question. The “Satisfaction categorization” row shows the computed total average:

Figure 3 - Categorization table

In the event a categorization is added to a parameter, then an additional Calculation Rule option becomes available:

• Avg. of Answers (the average of the answers) - this is derived from the sum of all the individual answers divided by the number of individual answers.

Note: All of the Calculation Rule options will give the same result if all respondents have answered all the questions. However the results will differ if there are null values (some respondents have not answered all the questions) or elements in the answer list have not been assigned weights. So which option you use will depend on whether:
- You want all respondents to have equal "weight" regardless of how many questions they have answered.
- You want all questions/statements to have equal weight regardless of how many respondents have answered them.
- You want all responses to have equal "weight".