When using multiple choice question type, only checked checkboxes are saved in the data set (value "Yes [Y]"). This leads to the problem that when analyzing multiple choice questions (e. g., with R or SPSS Statistics) you can only count how many people checked a response. But when analyzing a multiple choice question you always need the number of people who saw a question but did not check a response (e.g., percentage of yes vs. no). The current implementation becomes really problematic if you have to deal with incomplete data sets (which is almost always the case) or when subquestions are not shown to all respondents (because of filtering/conditions).

Imagine a survey with three multiple choice options. SQ1 is visible for all participants, SQ2 is conditionally visible only if SQ1 was checked and SQ3 was visible for all.
Now we have four participants:

• id 1 checked Q1, saw Q2 but did not check it, and checked Q3
• id 2 checked Q1, saw and checked Q2, but did not check Q3 (item non response)
• id 3 saw Q1 and Q3 but quit participation on that page (break-off)
• id 4 saw Q1 and Q3, did not check any of the options but submitted the survey (lurker)

The current data structure looks like this:

id | Q1_SQ01 | Q1_SQ02 | Q1_SQ03
1 | Yes [Y] | | Yes [Y]
2 | Yes [Y] | Yes [Y] |
3 | | |
4 | | |

So when just looking at the data set, you don't know which respondent saw which subquestion and chose to answer it or not. This really makes analyses and reporting very difficult and can easily result in wrong interpretation of data. Especially id 3 is hard for analyses (keeping real nonresponse and no answer because a subquestion was not shown apart).

To avoid these problems, the data set must look like this:

id | Q1_SQ01 | Q1_SQ02 | Q1_SQ03
1 | Yes [Y] | No [N] | Yes [Y]
2 | Yes [Y] | Yes [Y] | No [N]
3 | | |
4 | No [N] | | No [N]