Q-BANK: Development of a Tested- Question Database Questionnaire Design Research Lab Office of...
-
Upload
phillip-kennedy -
Category
Documents
-
view
214 -
download
0
Transcript of Q-BANK: Development of a Tested- Question Database Questionnaire Design Research Lab Office of...
Q-BANK: Development of a Tested- Question Database
Questionnaire Design Research Lab
Office of Research and MethodologyNational Center for Health Statistics
Cognitive Interviewing
at NCHS
A Brief Overview
Questionnaire design problems
Some possible threats to accuracy derived from questionnaires: Questions not understood as intended Don’t adequately capture respondent
experience Pose an overly challenging response task Don’t provide a legitimate option for an answer
Questionnaire design problems (2)
Errors can be systematic (bias) or random (noise; high variance)
Problems may not be visible in the actual data, inviting delusions about the quality of self-reports that have been collected
“In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward
“In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward But suppose we ask:
What do people think of as their “abdomen”?
What, to you, is your abdomen?
“In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward But suppose we ask:
What do people think of as their “abdomen”? What sort of threshold do people use for
answering “yes”?
“In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward But suppose we ask:
What do people think of as their “abdomen”? What sort of threshold do people use for
answering “yes”? What period of time are people thinking about
while answering?
“In the last year, have you been bothered by pain in the abdomen?” Possible revisions:
Show shaded picture of abdomen Drop “bothered” Use “In the past 12 months”
“In the past 12 months, have you had pain in the abdomen? By abdomen, we mean the shaded area in this picture.”
Clear alternatives address these problems, with no apparent drawbacks
Shaded picture of the abdomen
What is cognitive interviewing? 1) An administration of a draft
questionnaire…2) With additional verbal information
collected about the responses…3) And used to either:
Evaluate the quality of responses Or help to determine whether the question is
generating the sort of information its author intends.
Within these general parameters, there are a number of variants:
Practices of cognitive interviewing The “purist” version– encourage
participants to think out loud while answering questions; interviewer “prompts” participants without much other activity.
Practices of cognitive interviewing (2) Newer versions rely much more heavily on
direct probing. But even so, there can be much variation: Probes scripted or based on
interview/participant particulars (more “active” role for the interviewer)
Probes that are “cognitive” or more narrative in nature
A relatively realistic administration of the questionnaire (with debriefing at the end) or intensive discussion throughout
Some general practices We generally opt for:
Designing a semi-scripted interviewing protocol. It asks some probes in anticipation of potential problems, but allows improvisation.
Selecting a group of interviewees who would be typical recipients of the questionnaire and collectively cover a range of situations. Get a demographic/educational mix as much as possible.
Considering the interviewers to be investigators and not simply data collectors (methodologists with subject matter knowledge).
What sorts of probes should you ask?Consider this simple model of the response
process as a starting point:
Comprehension: Respondent interprets the question
Retrieval: Respondent searches memory for relevant information
Estimation/Judgment: Respondent evaluates and/or estimates response
Response: Respondent provides information in the format requested
Probes in cognitive interviews
Traditionally the majority of attention focuses on comprehension
Direct comprehension of terms: What does 'dental sealant' mean to you?"
But we also find that the most efficient way to learn about frame of reference and interpretation is to ask for narrative about the basis for the respondent’s answer Why did you answer that way? You said ___; tell me more about that.
Or, ask for paraphrase of the question.
Probes in cognitive interviews (2)To get at other aspects of the response process: Recall: How did you figure your answer to that? How
do you remember that? Confidence: How certain are you about that? (can be
used to identify judgment issues/uncertainty) Response categories: How did you pick an answer to
that? Was it difficult to find a response category? You can’t go wrong with:
“Tell me more about that?” “How so?”
The end result Qualitative analysis follows one or more rounds of
interviews Project reports include rich data on
Participant experiences and how these line up with question wording
Participant interpretations of the questions Problems participants had answering Potential ambiguities, trade-offs involved in asking
question a certain way Reports provided to sponsor, including
recommendations for changes when appropriate Then what?
Q-BANK Origination 2001 QUEST conference
Losing knowledge Repeat testing
Lack of “place” Implications:
Personal Administrative Survey Research
Impact on Question Evaluation Inability to compare studies
Hinders methodological conclusions Prevents insight Stagnant method
Limited use to survey research
Database Structure National Center for Health Statistics
original developers1. Access to past evaluation projects2. Methodological research in question design
Q-BANK houses: Individual questions Test reports
Test findings Methods used
Question characteristics
Question Subject Matter Numerous, open categories Accessed by drop-down list Examples:
Demographic: Income, Race, Gender, NationalityHealth: Insurance, Chronic conditions, SmokingEmployment: Status, Occupation, Job change,
Salary, Benefits
Question Information Type1. Events, Actions and Behaviors2. Objective Characteristics3. Subjective Characteristics4. Speculations5. Attitudes/Opinions6. Knowledge Tests7. Other
Response Category Type
1. Yes/no2. Select one textual3. Select one numeric4. Mark all that apply5. Allocation6. Open-ended textual: Delineated & Non-
delineated7. Open-ended textual: Delineated & Non-
delineated8. Other
Response Error Indicator
1. Interviewer difficulty2. Visual design problem3. Missed instructions4. Unknown terms5. Ambiguous concepts 6. Overly complex7. Assumption8. Double barrel9. Questionnaire effects10. Recall & estimation error11. Inadequate response options12. Problematic answer space13. None as tested
Value of Q-BANKSurvey Research Tool for:
Questionnaire Designers Subject Analysts Survey Methodologists
Interagency Effort Co-Sponsors:
Office of Behavioral and Social Science Research,
Bureau of the Census, National Science Foundation, National Cancer Institute, Bureau of Labor Statistics
NCHS provides internet access
Current Status On-line to steering committee members Testing and refinement Broader access possibly by end of year Looking for sponsors and partners
ChallengesPast Integrating interagency needs
Additional fields for establishment surveys Additional fields for self-administered surveys Reworking existing categories: response error
typologies, question type typologies, question topic categories
Establishing procedures Methodological requirements of evaluation
reports
ChallengesCurrent Coding consistency Populating the database Staffing limitations Funding limitations
ChallengesFuture Maintaining a methodological research
agenda Continual eye toward development
Searching in Q-BANK- Agency Name- Survey Title- Response Error- Advanced Searches
Q-BANK Search Fields
Agency Name Search
Agency Name Search
Agency Name Search
Q-BANK Reports
Searching on Survey Title
Searching on Response Error
Advanced Searches
Advanced Searches
Cognitive Testing Final Reports Don’t get lost in the weeds Important information about a tested
question is lost without referencing the final report
Conclusions about specific questions can be distorted without the final report
Contents of a Typical Report Introduction Method Procedures Summary of findings Question-by-question analysis
Why refer to the final report?
Search categories may not be straightforward Example: “At this time, is (CHILD) covered by
health insurance that is provided through an employer or union or obtained directly from an insurance company?” Type of response error = double barreled question
Why refer to the final report?
Search categories may not be straightforward Example: “At this time, is (CHILD) covered by
health insurance that is provided through an employer or union or obtained directly from an insurance company?” Type of response error = double barreled question
The interpretive process and underlying construct can’t be gleaned from the search category Example: “Have you ever taken any herbal
supplements for yourself?” Type of response error = ambiguous concept
Final Reports Q-Bank is a good tool for quickly finding
relevant reports of different cognitive testing projects
Searches not meant to replace final reports
Questions?
Kristen [email protected]