In this video, let's talk about how to best enter into SPSS multiple-choice item type
data.
First, let's be clear about when to use multiple choice items.
Imagine you have created the following question for your survey.
And this is an actual example from a student research project.
This student is asking -- "When you book hotel rooms for holidays, which booking channels
do you use?"
Then, he offers three options -- website of the hotel, travel agencies, and booking sites.
Before we talk about how to enter data collected from such an item, let's talk about whether
or not this is a good item.
My suggestion is -- no.
I would revise this survey question into Likert scale items.
Two things I'd change.
One: I'll change the 'you' phrasing into an 'I' phrasing, and I'll change multiple choices
into Likert scales.
So, instead of saying "When you book hotel rooms for holidays, which booking channels
do you use?"
I would say "When I book hotel rooms for holidays, I tend to use the following booking channels".
Then, I list out the options, each with a Likert scale.
In this way, not only do we collect data with more gradations and hopefully accuracy, this
also offers more possibilities in data analyses, AND this eliminates potential complications
in entering multiple choice item data.
Likert scale items are more straightforward to enter into SPSS.
So, in SPSS, this survey item will become 3 Likert scale variables -- each representing
one booking channel.
All right, so this is the first point I want to make here -- before we actually use multiple
choice questions for the survey, let's take a look at these questions and think about
whether multiple choice is really the way to go.
Many times, it is better to actually convert this multiple choice question into Likert
scale items.
There certainly are situations when multiple choice questions are needed.
One of those situations is this -- for example, I want to survey students and ask them -- Which
analytical method should be used to test the relationship between age and salary?
I offer 4 choices -- t test, correlation, chi sq, and SD.
Obviously, in this case, the only correct answer or at least the best answer is correlation.
So, how do we enter these data?
We go to SPSS, we create a new variable, let's call it Q1 as in test question number one.
It is a nominal categorical type variable.
It's got four options, let's code them -- 1 is t test, 2 is correlation, 3 is chi sq,
and 4 is SD.
Now, let me make up some random data as if we collected a lot of data.
Let's say we now want to find out how many people answered this question correctly, we
can easily find out by going to analyze, descriptives, frequencies.
Now, in the output table we can see, out of 60 people, 30 gave the right answer of correlation.
So, this is how we can code and enter multiple choice item data when the respondent is supposed
to choose only one out of several multiple choices.
Now, let's take a look at an example where respondents can choose more than one option
from a list of multiple choices.
Let's imagine another test question which asks -- Which statistics can be used as a
measure of central tendencies?
The answers are -- mean, mode, median, and range.
Clearly, the first three are the correct answers.
In this case, as we do expect that respondents will choose multiple answers from the four
available answers, we should code the variables differently from the previous example.
We are going to create 4 variables instead of just one.
So, each of the four multiple choice items will become a variable.
So, mean, mode, median, range.
We code the values with 0 and 1, with 0 representing not selected and 1 representing selected.
Once again, these are nominal measurement level.
Now we go back to data view, I am going to make up some random data.
To analyze these data, we can simply do a frequency on these variables.
Analyze, descriptive, frequencies.
In the output, we can see how many people chose each of the four options.
Another thing we might want to do is to see how many people gave precisely the right answer.
In other words, we want to see how many people chose mean mode and median but did not choose
range.
There are different ways to do that, I will show you what I consider to be an easy and
straightforward way to achieve this.
We use 'select cases'.
Go to data, select cases, and we can tell SPSS the criteria -- mean is 1, and median
is 1, and mode is 1, and range is 0.
Then we simply count the frequency.
Now we can see in the output very straightforwardly -- out of 60 people, 29 gave the right answer
to this question.
One more thing I want to talk about is the 'multiple response set'.
Sometimes people may be under the impression that if we have multiple choice data, we must
use multiple response sets.
So, let's take a look at how it works.
We go to analyze, multiple response, we define the set by choosing the multiple choice variables,
count 1, name it, and okay.
Now that we have defined the multiple response set, we can run frequencies, so go back to
analyze, multiple response, frequencies.
We see the output table with frequencies on all four variables.
We also use the multiple response set to run crosstabs.
We could even go to custom table and use the multiple response set to get additional statistics
such as the t, the z, and again, Chi sq.
But the thing I want to mention here is that all these analyses by multiple response sets,
whether it is frequencies, crosstabs, t, z, Chi sq, all these things, they can also be
obtained through descriptive analyses manually.
So, if you are a bachelor student researcher who is not very experienced with the multiple
response set, you don't have to use it.
Stick with the descriptive analyses that you have learned and practiced during your education,
and just use them.
You would be able to get basically all the necessary outcomes through descriptive analyses
as you would get from multiple response sets.
Finally, I am going to end with a little bonus fact.
Is it pronounced Likert or Lai-kert scale?
I always say Likert but some of my colleagues always say Lai-kert.
So, just to clarify, it SHOULD be pronounced Likert, because the psychologist who invented
this scale is a gentleman named Likert.
That's why I say -- Likert scale.
Thanks for watching, please like and sub, good luck with your research, and see you
next time.
No comments:
Post a Comment