In Magpi, checkbox questions are treated as a set of yes/no questions (this is different from the way they were treated in EpiSurveyor). So, for example, if you have a checkbox question like this:
You can really think of this as three questions:
Are apples one of your favorite fruits? Yes/No
Are mangos one of your favorite fruits? Yes/No
Are bananas one of your favorite fruits? Yes/No
When you analyze the data from this question, even though the question itself is designed as one question, and is displayed on the phone screen as one question, in the dataset it is displayed as three yes/no questions (ie columns) to make it easier to analyze. Each of the three columns will have a name created automatically from the data field name ("fruits") and the option name (e.g. "apples").
So if someone responded to this question by checking "apples" and "mangos" but leaving "bananas" unchecked, it would display in the dataset like this:
This makes it much easier to analyze the data from checkbox questions than in EpiSurveyor.