I have a set of data whereby one of the questions states: How many foods from the following list do you eat on a weekly basis. Tick all that apply. There are 8 possible answer options which include: Brown bread, an egg, an apple, ice cream, bacon, a chocolate bar or fish.
My question is, how can I analyse the data if say I want to find out what factors predict that someone will eat brown bread on a weekly basis? I have explored some options which revolve around generating dummies for all possible answer combinations but this would be impossible!
Any ideas?
Dear Georgina,
This is known as a multiple answer multiple choice question. Each answer generates a separate binary (yes/no) variable so you would end up with 8 binary variables.
If you want to try to predict one of the variables from the others then you can use binary logistic regression. This works with binary independent variables and there is no need to use dummy variables.
If you want to compare the relationship between the answers then you can use Cochran's Q.
Both these tests are available in SPSS.
I don't see any problem with creating dummy variables for each of your answers, because they do not overlap. That is, eating brown bread generates a response that is separate from the response to eating ice cream, etc.
If your are not committed to doing regression, you might consider cluster analysis and multi-dimensional scaling to find patterns in which founds are eaten together.
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design