Archive for May 2nd, 2006

How to handle multiple response questions

In a lot of research, multiple answers can be given to a single question. For example:

“What kind of food do you like?”
o soup
o rice
o salad

How do you analyse this type of question? Let’s assume you want to make a table with the answers. How do you combine them? Even though it may seem like the most easy thing to do, this is pretty difficult stuff. SPSS is good at analyzing unique combinations of variables (answers to questions) combined with unique cases (people in a survey). The combination of more than one answer per person, does not fit into that logic. So we have to be a little creative.

There are – at least – four different ways to analyze these results. They all have advantages and disadvantages, so it really comes down to your preferences:

1) Create separate variables for each answer 
varsoup (0 = not ticked, 1 = ticked)
varrice (0 = not ticked, 1 = ticked)
varsalad (0 = not ticked, 1 = ticked)

If you’re not entering the data yourself, chances are that this is what your data-set will look like if there was a multiple question in the survey. With three separate variables, you can create three tables using the frequencies-command. The advantage of this approach is its simplicity, the disadvantage is that you have three separate tables. You cannot tell if the three answers were in any way linked to one another.

2) Make a ‘grouping variable
a. Follow the steps in option 1, creating separate variables for each answer
b. Create a new variable: the ‘grouping variable’. Select analyze > tables > multiple response sets… Select all variables from the list you want to group together and click them into the right window ‘Variables in set’. Choose the ‘counted value’: this is the value that you want to count as ‘yes’. In our example, this is the value 1. Give the new variable a name, for instance ‘varfood’. Click ‘add’, and ‘OK’. Your new grouping variable will appear with a $ in front in de list to the right of the screen: ‘$varfood’. The $-sign tells you this is a variable containing several variables.

To make a table, select analyze > tables > multiple response tables… In the window to the bottom-left of the screen, you see grouping variables. Click the one you want and click it into the ‘Rows’ box. Click OK and you’re done.

3) Several variables with a hierarchy
In order to be able to use this method, it would be nice if your survey also asked to rank the three items available. If that’s not the case, you will have to decide for yourself which answer is most important and label that one first answer.

This will lead to:

varfood1 (1 = soup, 2 = rice, 3 = salad) label “First answer”
varfood2 (1 = soup, 2 = rice, 3 = salad) label “Second answer”
varfood3 (1 = soup, 2 = rice, 3 = salad) label “Third answer”

4) Create a variable with a single value for each possible combination
This will lead to:


1 = soup,
2 = rice,
3 = salad,
4 = soup and rice,
5 = soup & salad,
6 = rice and salad,
7 = soup and rice and salad

The disadvantage is that this can be quite some work when you’ve got more than three answers. And you also run the risk of not being able to interpret your resulting table at a glance: the number of cases per option can be quite small. This is really only a serious option if you want to know exactly what each person answered. Pay attention to the order of the answers: if you start off with those combinations that were pretty popular, they will be at the top of your table, making it easier to interpret the results.

(Thanks to Sander for answering this question)

27 comments May 2nd, 2006


May 2006
« Apr   Aug »

Posts by Month

Posts by Category