Working With Evaluation Sets
Evaluation set is a combination of two or more questions. It is important to remember that each respondent answers these questions more than once. For example there are separate answers to set of questions "Brand" + "Is reliable" for every brand - 6 answers.
It is possible to use only one evaluation set in a particular query. This means that if you want to include more than one evaluation set in a project you have to put them in separate tabs.
Example
In the sample data there are two evaluation sets available - 'Brand image' and 'Advertising awareness and performance'.
Variable 'Brand' from the first evaluation set is not the same variable as 'Brand' in the second evaluation set. They contain different data.
|
Brand image
|
Advertising awareness and performance
|
|
|
|
It is impossible to analyze two evaluation sets in one query that is why they are placed in separate tabs. The results are presented below.
In these two questions respondents are evaluating different brands (not only the make of phone that they bought). If you want to see how users of make A evaluate all brands you have to include filter on variable 'Make of new phone' (not on 'Brand'). See results below.
Using Evaluation Sets In Filter
Questions which belong to evaluation sets can be used inside Filter expressions, like any other questions. However, the way how the filter works depends on other elements in the query.
Case 1:
If the query itself (the calculation and other questions by which the output is sliced) does not reference the evaluation set then the filter expression simply limits respondents (cases) to the ones which match the filter condition.
Example 1:
If you have an evaluation set containing responses about "Advertising awareness" and use "Brand" question from this set into the Filter expression limiting it to "Brand A" then the calculation for e.g. Average age will return the value calculated within all respondents who gave an answer to "Advertising awareness" for "Brand A".
Case 2:
If the query is referencing other questions within the evaluation sets, the filter will limit the calculation to specific evaluations which match the filter condition.
Example 2:
If you have an evaluation set containing responses about several "Image" aspects for a given "Brand", and you apply filter based on this evaluation set for: "Brand=Brand A AND Is reliable=1" then subsequent calculation for mean score of image statement "Overall brand image rating" will return average scores for this question only for "Brand A" and only for those respondents who gave "Brand A" a score 1 in "Is reliable".
In a specific scenario you may want to override the default behavior of the filter expression described above in Case 2 by forcing the outcome of the Filter operation to always be respondents (cases), not evaluations. You can achieve this by encapsulating the filter expression inside a UDF and setting the filtering mode to "Match respondents" (in contrast to the default mode which is "Match evaluations"). For more details see Filtering modes for expressions with evaluation sets.
