IntroSurveyResearch - D-Lab

advertisement
Survey Research Methodology
Dlab Workshop
Leora Lawton
February 9, 2015
1
Outline





Purposes of survey research
A brief foray into sampling
Considerations in design
Phone, Mail, Internet, Face-to-Face
Incentives
2
Surveys Defined






A technique for gathering data, recording them with numeric
codes, for the purpose of statistical analysis.
Surveys are quantitative but can have qualitative components to
them.
Surveys typically involve interviewing individuals using mail,
internet, face-to-face or telephone methods, but may also involve
recording of existing data into a quantitative data base.
Require identification of the research goal, sampling design,
questionnaire design, analysis plan, and reporting.
Survey research rests on a long history of scientific research for
optimizing data quality through minimization of sources of
statistical error. They are part art, but they are definitely science.
Just because you’ve taken surveys does not mean you know how
to design or conduct them.
3
The Process
1.
2.
3.
4.
5.
6.
7.
8.
9.
Identify the research objective and verify that it is best served with
a quantitative approach. Create an outline or conceptual model of
your objective.
Select the design most suited for your objective.
Keep in mind that preliminary qualitative work may be required to
identify survey items (questions).
Identify the population best suited for providing the information
and locate a list of potential respondents (sampling frame).
Identify the means best suited for reaching this population.
Design the instrument (questionnaire) using the conceptual model
as your outline. Review, revise, test (repeat)
Collect the data
Prepare the data set for analysis
Analyze and report.
4
Defining the Objective
•
Before you know what to write, you must first figure out what you want to
know.
Quantify
Describe
Correlate
Categorize
Explain
Causality
Determine
Predict
Classify
Track &
Trend
Changes
Example
• Problem: There is a need to roll out a new public health program.
• Research Objective: How can we design a program that will meet the
public health goal and also resonate with the culture of the population we
are targeting.
5
Identify the Relationship: I

The conceptual model and framework is developed in your dissertation proposal.

A conceptual model consists of the relationship between that which you seek to
explain (which will be the dependent variable(s)) and the explanatory factors
(independent variables). The relationship is explained by the theory.

Example: An education program results in more students continuing to grad school
Support,
Value
Meaningful Experience
Enjoyment
Enrolling in
Grad School
6
Identify the Relationship. II

Next, you begin to identify the detail of factors that will be used in your survey as real
questions that people can answer meaningfully, with validity and reliability.

Example: To build trust, one uses locally hired and trained staff, so it has familiarity
and also adds more resources to the environment. The location of the service will be
in the already utilized local health clinics. This will lead to more use of the service.
Mentoring
Stipends,
scholarships
Peer
Relationships
Intention,
Applications,
Enrollment
Classes,
Conferences
7
From Objective to Design
• Once the objective is clearly identified, then you
can begin thinking about
– Sample –
• Who can provide you this data,
• Who is involved in the process or phenomenon,
• Do you need a control or comparison group?
– Design
•
•
•
•
Experimental
Cross-sectional
Panel
Longitudinal
8
From Objectives to Design
• Experimental: Refer to Shadish, Cook and
Campbell if using an experimental design
– Comparing groups (control vs test)
• Randomly assigned
• Not randomly assigned
– Self-controls – before after designs or panel (same
respondents
– Historical – comparing results from one sample with a
later one (different respondents)
9
From Objectives to Design
• Descriptive
– Cross sectional (snapshot)
– Cohorts – same people over time, or same
sample frame over time (longitudinal in my
book)
– Case control - matching
10
Sample and Sampling
• Who do you need data from?
• Will they be able to provide meaningful
data?
• And how can you reach them?
Perhaps the biggest challenge in data
collection is identifying a target population
that you can, in fact, reach.
11
Sample
• Universe – target population
• Sampling frame – list of people in your universe
• Sample – the selection of potential respondents
to be contacted
• Respondents – the units who actually provided
data
– Non-respondents – the ones that didn’t.
– Item non-response – didn’t answer specific questions.
12
Sampling
• Once the sample population has been
identified, you need:
– Sample size
• See Sample Size Calculator in Resources section
of class website
– Sample frame creation and selection
• Coverage
• Determination of eligible sample frame elements
• Sources for acquiring samples
13
Sampling
• Sampling strategies
– Probability samples: each element has a > 0
probability of being selected
•
•
•
•
Simple Random Sampling
Cluster sampling
Stratified
Multistage
– Non-probability – Snowballs.
• Respondent Driven Sampling (RDS)
14
Data Collection Methods
and Survey Instruments
• Methods – Instruments
–
–
–
–
–
Face to face – paper, CAPI, handheld
Telephone – paper, CATI, IVR
Internet – email, web-based
Mail – paper, scan-form
Fax – paper
Also…
– Mixed Mode – e.g., fax, web, mail
– Multimode – phone-mail-phone
15
Data Collection Method
and Survey Instrument
• Each method has advantages and
disadvantages in terms of:
– Ability to reach correct target
– Budget
– Time
– Error (non-response, sampling error)
– Appropriateness of data needed
16
Data Collection Method
and Survey Instrument
• Face to face:
– Less commonly used
– Often residential, door to door
• Mall intercepts
• Executive one-on-one in-depth interviews (IDIs)
– Very time consuming and expensive
– Well-trained interviewers, often matched for ethnicity,
race and/or language
– Residential sampling often uses census blocks
– Paper, CAPI, handheld, or multi-mode.
17
Data Collection Method
and Survey Instrument
• Telephone:
– RDD now challenged by lifestyle and tech
developments
– Both business and residential
– Moderate time, but fairly expensive
– Trained interviewers, may match for ethnicity,
race and/or language
– RDD, listed sample, or specialized lists
– CATI, paper
18
Data Collection Method
and Survey Instrument
• Internet:
– Replacing much telephone and mail research but
losing representativeness. Ease of fielding introduces
design and measurement error.
– Both business and residential
– Quick data collection, often at a very low cost
– No interviewer effects, but could be browser effects
– Specialized lists or ‘pop-ups’
– Web-based ASP, software, embedded forms.
19
Data Collection Method
and Survey Instrument
• Mail:
– Some predict a come-back.
– Both business and residential
– SLOW data collection, moderate cost
– No interviewer effects. No control on
respondent
– Specialized lists, census-based lists.
– Paper, Scantron or similar forms.
20
Special Populations
• Children
– 13-17
• Similar to adults, but make the language more conversational
and perky. Can use icons for survey rating scales.
– 10-13
• Short, with pictures, but some abstract thinking, and they
have already started taking assessment and similar tests, so
or more familiar. Can look like survey
– 6-9:
• short, with pictures. Children are VERY literal. Limited
abstract thinking, variable reading skills. Make it a game.
• If web-based or CASI, add voice-overs.
– 5 yrs or under:
• do ethnographic only
21
Response Rate (brief intro)
Your goal in research design is to reduce error as much as
possible and enhance response rate as much as
possible.
• Response rate =
– # responded / (# potential approached – ineligibles)
• Non-response = 1 – RR
• Examples of non-response:
–
–
–
–
–
Item non-response
Terminations
Refusals
Language
Callbacks
22
Writing a Questionnaire
• The Groves et al. chapter focuses on the
principles of asking questions, that is, the
cognitive and psychological processes around
respondents’ reading, interpreting and
answering questions (also known as ‘items’).
• The Dillman chapter (2) talks about specific rules
around the items.
• Also, look at the Questionnaire Guidelines
document (.pdf) on the website for an overview
of the process and the most common guidelines.
23
Writing a Questionnaire
Your objectives are the ‘map’ to the actual
questions.
– Which factors in membership organizations identify
oligarchy?
– From literature review:
•
•
•
•
•
•
Communication to members
Input from members
Following by-laws
Leadership cycles out of office
Stability of executive office infrastructure
Crisis
24
Writing a Questionnaire
• Further break down the concepts until you can come up
with specific issues around which questions can be
written.
– Mentoring
•
•
•
•
Did you visit the GSI consultant in Dlab (yes, no)
How often?
Did you work with any faculty members?
Describe that experience (open-end)
– Financial support
• How important was the stipend for you?
• How would you have paid for tuition if you didn’t have this program?
– Classes
• Which courses did you take? Overall, how much did you think the coursework
contributed to your ability to do research?
– Conferences
• Did you attend an annual meeting? Did you present a paper? What else did you do at
the conference
– Peer relationships
• Did you make friends in this program?
25
Cognition
• Read/hear > Interpret > Identify answer >
Provide answer
• At any stage, respondent burden can overwhelm
the process.
• At any stage, researcher error can put up a
stumbling block or hurdle.
• At any stage, the survey instrument, whether
interviewer-administered or self-administered,
can block the process.
26
Respondent Burden:
Don’t make them work
• Read/Hear: The question is too complex,
wordy, sloppy.
• Interpret: Grammatically incorrect, outside
the flow of thought, backwards logic, faulty
logic, unfamiliar terms.
• Identifying answer: Meaning to item
found, remembered, calculated. Or not.
• Answer provided: correctly or incorrectly,
social desirable, satisficing.
27
Critical Feature
• Once the data are collected, you must be able to
answer the questions specified in your
objectives. You won’t be able to if you don’t go
through the objectives clearly, and map the
questions to them.
• If you don’t know what you are going to do with
the data once it’s collected, you may not have
the right question, or you haven’t been following
objectives, or have none identified.
– The question is unnecessary
– The question is off-target
28
Critical Feature
• Examples:
– 1. Question:
• How satisfied are you with the ease of reaching a
representative by telephone?
• What’s the objective??
• Possible meanings:
–
–
–
–
Time to answer phone (number of rings)
Number of ‘buttons’ to push before reaching correct rep
Amount of time on hold.
Number of reps before reaching correct one.
– 2. Objective:
• Understand interest in viewing edited reruns
• Determine likelihood of viewing on another cable network
– Question: How likely are you to watch an edited program on
another network?
– Will this question work? Why or why not?
29
Getting on first base
• Because first impressions are all you have
before they toss in the mail, hit the delete
button, or hang up.
• Think carefully about how to reach your
target respondent. Especially for
organizational studies (or businesses) you
may need to try a handful of more
qualitative approaches to see what works
best.
30
Telephone Introduction: Principles
• Phone call
– Gate keepers and screeners – probe for correct
respondent.
• Busy > call back (6+ times)
• No Answer > call back (6+ times)
– Alternate time of day, day of week (can program this
into CATI)
•
Refusal: Soft > ask to reschedule or call
another time. Employ different interview who
has good track record with converting refusals
• Refusal: Hard > apologize for intrusion. Ask for
another time, another person. Code as refusal.
31
Telephone Introduction: Principles
•
In our introductions, we:
•
* introduce ourselves - interviewer's name and Indiana University Center for
Survey Research;
* briefly describe the survey topic (e.g., barriers to health insurance);
* describe the geographic area we are interviewing (e.g., people in Indiana) or
target sample (e.g., aerospace engineers);
* describe how we obtained the contact information (e.g., the telephone number
was randomly generated; we received your name from a professional organization);
* identify the sponsor (e.g., National Endowment for the Humanities);
* describe the purpose(s) of the research (e.g., satisfaction with services
provided by a local agency);
* give a "good-faith" estimate of the time required to complete the interview (this
survey will take about 10 minutes to complete);
* promise anonymity and confidentiality (when appropriate);
* mention to the respondent that participation is voluntary;
* mention to the respondent that item-nonresponse is acceptable;
* ask permission to begin.
•
•
•
•
•
•
•
•
•
•
32
Telephone Intro: Example
•
"Hello, I'm [fill NAME] from the Center for Survey Research at Indiana University.
We're surveying Indianapolis area residents to ask their opinions about some health
issues. This study is sponsored by the National Institutes of Health and its results will
be used to research the effect of community ties on attitudes towards medical
practices.
•
The survey takes about 40 minutes to complete. Your participation is anonymous and
voluntary, and all your answers will be kept completely confidential. Your telephone
number was randomly generated by a computer. If there are any questions that you
don't feel you can answer, please let me know and we'll move to the next one. So, if I
have your permission, I'll continue."
•
At the end, we offer the respondent information on how to contact the principal
investigator. For example:
•
"John Kennedy is the Principal Investigator for this study. Would you like Dr.
Kennedy's address or telephone number in case you want to contact him about the
study at any time?"
33
Mail Intro: Principles
• Dillman’s Tailored Design Method
– Advance letter, on letterhead, from sponsor
(where applicable)
– Packet contains cover letter, instrument,
postage-paid return envelope, $2 incentive.
– Postcard reminder 2 weeks later
– Second packet 4 weeks later
– Follow-up phone call if possible
34
Email Intro: Principles
• Use principles of social psychology
–
–
–
–
–
–
–
Politely greet
Ask for help
Appeal to altruism
Anchor to self-interest
Define help needed
Tell them what to do
“K.I.S.S.”
– “From” from identifiable source, personally known (not a
personal acquaintance, but where a relationship exists)
– “Subject” Avoid the look of spam.
– Follow-up with reminder emails. Most responses occur with 72
hours. Follow-up can be 4-7 days later. Too many follow-ups is
perceived as spam.
35
Email Intro: Examples
•
Email invitation
– From: An organization you know
– Subject: Take a survey about your <interest>
– Message: Dear <name if possible>
• We are conducting a survey to evaluate our recent annual meeting in Montreal. Please
take a couple of moments to share with us your experience and opinions. The survey
will take about 4 minutes, and will greatly benefit future planning in ORG so
conferences will be even more rewarding and enjoyable for you. Just click on the link
below to get started. (If no button appears, copy and paste the URL into your browser.)
• Sincerely,
• John Doe, Survey Director
If you have any questions, contact me directly at 800 555-1212, or email jdoe@org.org.
•
Pop-up
– Hi! We are conducting a survey to evaluate our website. You can help by
providing your opinions. Doing so will make this website better for you. The
survey is short: just 3 or 4 minutes. Please click on the link below and get
started!
– (top of survey) Thanks for taking our brief survey! We greatly appreciate all of
your feedback.
36
Incentives
• Few topics are covered in more depth than
the issue of incentives.
– How much for which audiences is necessary
to maximize participation?
– Do incentives cause bias?
– What form? Cash? Coupons? Gifts?
– When cash/gifts are not acceptable, then
what?
37
Incentives
• Mail surveys
– $2 more than $1, more than nothing. $5 not much more help.
– Some suggest higher incentives, upon second try, especially for hard-toreach populations.
• Phone Surveys
– Often not needed. Long ones should have it though.
• Email Surveys
– Experiments with prizes, drawings, cash, coupons, gift cards.
Rewarding panel members especially critical.
• Special Audiences – consider summaries of results, charity
contributions.
– Government
– Business
– Physicians, IT, other oversampled professions: More Cash. $75+
38
Examples of papers from
AAPOR 2006
• Increasing Response Rates with Incentives
– Non-Monetary Incentive Strategies in Online Panels
– Incentive Check Content Experimentation
– Effect of Incentives on Mail Survey Response: A Cash and
Contingent Valuation Experiment
• Experimenting with Incentives
– Are One-Time Increases in Respondent Fee Payments CostEffective on a Longitudinal Survey? An Analysis of the Effect of
Respondent. Fee Experiments on Long-Term Participation in the
NLSY97
– Personal Contact and Performance-Based Incentives
– Effect of Progressive Incentives on Response Rates
– Lottery Incentives with a College-Aged Population
39
Download