[BLANK_AUDIO]

Welcome.

So, in this segment we will talk about different types of questions that

you might want to design or will encounter when you deal with surveys.

I will give you some examples and some terminology so that you don't

get lost, when throughout the course, we

will talk about different types of questions.

This first one here is a simple question.

It has a YES/NO answer category.

One phrase.

No explanations, no special features here.

It is, in a sense, the most simple form of question that questionnaires usually have.

The examples I'm going to use come from "Monitoring

the Future" and so does the simple question here.

If you're, curious about that survey, you can

check out the website that comes with it.

It is a survey sponsored by the "National Institute on Drug Abuse".

Just part of the "National Institute of Health".

.. and it is conducted here at the Institute

for Social Research at the University of Michigan.

Another example from "Monitoring the Future" is this question here.

Which has a running prompt.

We call it that.

You see it's a question, there are a

couple of different sub-items as we call them.

This particular one, on how many

occasions, if any, have you used marijuana?

And then the first running prompt that we have is in your lifetime.

And, over here on the right, you see a couple of answer categories.

These can be different scales.

This is what we call a self-administered questionnaire.

It is very nicely formatted.

The layout is pretty clear so that people can

see where the answer categories are, where the labels are.

And throughout the course we will talk more about

how many scale points to choose in those kind

of scales and what labels one should be using,

how you should think about labeling this kind of questions.

The next one here is a closed question.

Now, the other ones were also closed, but they were much simpler.

Here, we actually have scale points like in

the one before, but it's a one choice question.

We have one question, how do you think your

own life will go in the next three years?

Do you think it will get better or worse?

And then a scale to pick an answer.

This is in contrast to questions like this one here that are sometimes called

"mark all that apply", or questions that allow for more than one choice.

And, this particular question here is, a question for race and ethnicity.

Sort of a mix that too will be a separate topic in this course and

you can see here that people are allowed and there are instruction here that mark

that to the respondent to select one

or more answer categories, so, someone could be

white or African American and white, easily imagined,

moreover, if you have parents of mixed race.

Here's another example for mark all that applies, this is

again a question from "Monitoring the Future" about drug use.

And the question here is what methods have you used for taking cocaine.

And, obviously, respondents can pick several or all of these answer categories.

One note ahead, this is not a course about data analysis,

but if you start converting this kind of questionnaire or this

kind of questions into a data set, it is important that

you already think ahead how to code this kind of question.

Because, later on in your data set, it

probably would show up as six different variables.

One variable 84.1 sniffing and snorting and then

a yes no, a zero one coded in there for this question having been checked or not.

If you don't pay attention to this,

then data analysis can be quite complicated afterwards.

I'll talk about analysis issues in a

separate segment on preparing your analysis plan.

Okay, so this question here is what is called a filter question.

You see here that the layout is very different.

And so let me show you first that this is a different source.

This is the British Crime Survey from 2002.

The new version is up there, but this one

is already in the archive and easily accessible for everybody.

This is an interview administered questionnaire.

So you see a lot of code that

the respondent doesn't see, but the interviewer does.

Or if it's automatically programmed into a computer aided interviewing device.

The program can pick up these code words.

Knows which questions these are, and the like.

What is key in this YES/NO filtering setting is that the

interview will read to you apart from anything you have already mentioned.

This is in brackets in case nothing was mentioned yet.

The question continues, in that time did anyone get

into your house flat without permission and cause damage?

So, capitalized words that the interviewer should pronounce.

And there is the YES/NO.

And here is the hint for the computer program to continue asking followup

questions to this filter question if this

previous question has been answered with yes.

So, this is what we call a filter question.

This has the advantage that respondents don't have

to say, oh this question doesn't apply to me.

When they are asked, it shortens questioners

for people, for example, in this case.

Never had any burglary at their house.

And it allows, therefore, tailoring of our questions to respondents.

Now an interesting side note here sometimes, and there has been research out

there, respondents do clue into this problem

and oh, into this question is structure.

And they might actually neg, negate filter questions.

They know here to avoid the follow up.

So this bit of research that I'm happy to report

in a separate segment here not particular part of the course.

But look out on the website and we'll show you results from that particular research.

And finally, there are what is called open ended questions.

Those question are questions like this one here.

This is from a customer survey that KTTC in the U.S. has issued.

Down here there's a question.

Please tell us what you like best about the news on KTTC.

And then there is an open field in this particular

self administered web survey where respondents can type in their answers.

Again, we have a separate segment which will talk

about that design of web surveys in more detail.

But as a small hint on the, the amount of answer open

ended text you will get does depend on how big that box is.

Respondents do react to that.

Okay, so in the next segment we will talk about bias and variance.

Those are the issues I just hinted at, you know, how do respondents vary their

answers as a result of how the question is asked, how the layout is created.

And what respondents might anticipate the question

meaning is and things of that nature.

Types of Questions

社会调查的问卷设计

Meet the Instructors