Please enable JavaScript.
Coggle requires JavaScript to display documents.
4: Developing a questionnaire (Q) (Wording items: Neutral statements vs…
4: Developing a questionnaire (Q)
Ways of collecting information
Self-administered Q
best for:
Variables with too many values/categories to read over phone
Non-observable attitudes/opinions
large population (in short time)
Research on stigmatizing behavior difficult face to face through anonymity
Generalizing to a larger population
Anonymity:
Could make it difficult to trust that respondents are who they say they are
Follow-up mails/cards, gift incentives etc. increases response rates
Possible problems
Being able to read and understand questions is vital
Computer-based: Access is critical. Varies between race, income, age etc
Closed-ended items
may not have an answering option that reflects your behaviour
Structured interviews
-
Face to face and telephone interviews
:
Unstructured, in depth
:
Ideal for exploratory research
One by one or
focus groups
Better response rates than self-administered = improves accurate generalization
Questionnaire =
interview schedule
Face to face
:
Interviewer and interview process may affect answers
Answers should be written as accurately as possible without adding or deleting anything
Possible to include self-administered sections in sealed envelopes
Takes time, smaller samples, expensive
Telephone
:
More impersonal than face to face
Mobile phones increase sampling problems
Content analysis
: Extracting, sorting and measuring key elements from
open-ended
questions
Control:
Interrater reliability
Behavior often more reliable than attitudes.
Behavior
in questionnaire only tells what people say they do. E.g. combined with
frequency
: How often do you go to church?
a)weekly
b) monthly
c) yearly
d) less than once a year
Conceptualizing
Could start by writing hypothesis and variables
At least one questionnaire item for each variable in hypothesis
Outline: Three main components:
behaviors, attitudes/opinions, demographics
Write topics/questions. Write as many as possible, discard later!
Design should not be crowded, too small letters etc.
Filtering
: Ask if respondents have an opinion about a research subject before launching
contingency
questions
Intesity measure,
Likert-scale
Wording items
: Neutral statements vs "I feel that..." may provide different answers
Common mistake: Treat group as unified in attitudes, e.g. "my family supports me when..."
Avoid
negatives
in questions
Mix
direction
, so that "agree" is not always the positive answer. Cuts down on
response bias
and
socially desirable
answers
Always avoid
always
and never use
never
Avoid
double-barreled items
(measures two things)
Avoid
leading
or
loaded
questions
If asking respondents to
rank
items, limit the number of items. Also this tells nothing of gaps between items ranked
Don't ask quiz-questions that respondents probably don't know, e.g."how many of your colleagues support gay marriage?"
Alternatives must be
mutually exclusive
(only one possible option for a given respondent)
Also
exhaustive
: All possible options provided
Ask to rate rather than rank, to know how strongly respondents feel about the items
Find out what people do rather than what they feel
Options should be mutually exclusive and exhaustive (include all options with no overlap)
Beware of forced choice
Ask what respondents can remember. Not what they did two weeks ago on Tuesday
Beware of response bias. Possibly include trap, such as a fake book title among books respondents can say they've read
Place questions about respondent at the end. They'll likely prefer easy tasks after tiresome reflections. (reverse in interviews, where early easy questions can serve as ice breakers
Branch questions if necessary, i.e. "if
no
go to question 4"
Automatically in online surveys
Most interesting questions at the start, to make respondents positive about completing
Questionnaire should be built logically according to subjects and ways of answering in order not to confuse respondents
Pilot test the survey, first yourself, then others