Have you ever taken a survey that asks you to "check all that apply"?

Some people may check one response, or none, or all of the responses. That may be

an interesting way to ask a question, but it can be a nightmare if you do not know

how to handle that kind of data in IBM SPSS Statistics software, we are going to

dive deeper into how to handle multiple response items in SPSS.

Some questions can have only one answer. We asked teachers "where is your school

located?" Because the school can be located in only one place, there can only

be one answer. Now, this is clean and easy to use. As a researcher, each respondent

can only choose one option, so the results will be percentages that add up

to 100%. But other questions allow for multiple answers. Using this powerful

strategy in a survey can create problems for us later on, so we want to plan ahead

on how to treat these multiple response items in SPSS. In this question, we ask:

" at your school what sorts of curriculum modifications

are available for the deaf and hard-of-hearing students?" However, we add

the caveat "check all that apply". A given respondent may check all of the options,

or only some of the options, or none of the options. Perhaps 68% of schools offer

modified course instruction, 51% offer alternative forms of coursework, 30%

offer extended exam time - and now we're way over 100% - so, how can we display

these findings in a way that makes sense of the data? When you use multiple

response items, remember this important tip: (in this case we're asking people

about the types of pets that they own) think about whether a possible answer

could be zero or none. If it is, then include that "none of the above" option, so

that if a respondent does not have an answer, the non-responses are not treated

as missing data, but rather are recorded as legitimate zero values. Anytime that

you include "check all that apply", you need to treat the answer options in a

very specific way. You will use the multiple response option in SPSS

for this example we will use the GSS2016.sav dataset. This is a very

useful real-world data set for the General Social Survey - a cross-section of

the American public from the year 2016 - provided by the National Opinion

Research Center at the University of Chicago. It is free to download and I

will include a link. We are going to be examining some multiple responses

related to how often people attended an art exhibit in the last year, and if they

did not attend an art exhibit, why not? Let's look at some simple frequencies.

First, so that we will have them as a reference, go to Analyze -> Descriptive

Statistics -> Frequencies. We will examine two questions: "did the respondent go to

an art exhibit in the last 12 months?" (that is a yes or no question)

and "What was the most important factor in NOT attending an art exhibit?" in which

they can choose only one response from a list. Move these two variables into the

variables box and click OK. We can see that roughly one-third of people

attended an art exhibit in the last 12 months. In the second question, those who

did not attend an art exhibit were asked to the single most important factor why not?

We see that the most common objections are Time and Cost, but that

only tells us about the most important reason. Some of the lesser important

reasons may still be important factors. By asking about only the most important,

we lose all of the information about the second, and third, and fourth most

important reasons. So, we may follow our question about the single most important

reason with a question in which we offer a list of reasons and ask people to

check all that apply. People might answer yes or no,

to any, all, or none of these items. We will use a multiple response set to

analyze these data. When you use the multiple response option in SPSS, you

need to specify one particular response as the response that will count. Now the

simplest way to do this is to code your data as 0's and 1's where 0 = no,

and 1 = yes. Only response 1 is counted as an affirmative response.

When we look at the value labels for these items, we see that only response 1

indicates that this was a reason for not attending an art exhibit.

All of the other responses indicate something else. In fact, responses 0, 8, & 9, are missing

values. Those missing cases will not be included in our final analysis.

Our first step, then, is to create a variable set out of the possible item responses.

Go to Analyze -> Multiple Response -> Define Variable Set... I'm going to expand to this

window so that we can see the variable names a little better. Here we will tell

SPSS which items need to be combined for our multiple response set. We will move

any variable that relates to a reason why someone did not attend an art

exhibit: it costs too much, could not find anyone to go with,

not interested, did not want to go to that location,

lack of time, too difficult to get there, and other reason. Move all of the

relevant variables into the Variables in Set: box. The variables will be coded as

dichotomies with the counted value of 1. All other responses including missing

values will be treated as not counting for this question. We will name this set

of responses something that will help us remember what it is: why did people not

attend an art exhibit? We will call it "Artless."

Click Add to create the multiple response set. Close. Now at this point, we might expect to see

our output window, but nothing really happens and that is because all that we

have done so far is to DEFINE our variable set. Now we need to tell SPSS

what to DO with it. We need a frequency table. Go to Analyze -> Multiple Response ->

and notice that we have two analysis options" frequencies or crosstabs.

We're going to use Frequencies. Move the Artless response set into the box: Table(s) for:

We are going to get a frequency table for the items. Click OK. In the

output, we are given the number of people who responded to each item, the percent

of responses, and the percent of cases. We are most interested in the" percent of

cases" because we want to know how many people chose each response, we would

interpret these findings as 28.2% of people cited cost

as one reason that they did not go to an art exhibit and 14.7%

did not go because they did not have anyone to go with. Not having the

time to go - or not making the time to go - was the reason given by 60.9%

of respondents. Of course, someone who said they did not have the

time to go, might also have indicated that they did not have the money, either.

But now we have a richer source of data to better understand these respondents'

motivation - or lack of motivation - for attending art exhibits. One thing that I

DON'T like about this table is that it's difficult to read and interpret.

These data would be better displayed with a Pareto bar chart. To display these

findings, we're going to switch over to Excel and

create this bar chart. If you want to learn the steps to do that, check out the

next video.