You are on page 1of 21

Second Workshop for Improving Disability Statistics and Measurement

Bangkok, 27-29 September 2004

Questionnaire Testing and Interview


Techniques

Survey Cycle

Client Liaison

Evaluation Planning
Question and
instrument
Survey development and
Dissemination testing
Development

Survey Sample Design


Analysis

Estimation Data Collection

Processing

1
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Minimising measurement error


zRemember
ƒTotal survey error is the sum of sampling error

and non-sampling error


ƒSampling error is minimised through increasing

sample size
ƒNon-sampling error is bias and is minimised

through attention to survey procedures, such as,


development, questionnaire, sample design,
interviewing and processing
zAim of testing questionnaires is to minimise
measurement error

Sources on non-sample error

2
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Questionnaire Testing
zTwo broad categories used
ƒPre-Testing
ƒField Testing

Pre-test properties
zOccurs:
ƒBefore field tests

ƒSmall, targeted sample

ƒControlled environment

ƒQualitative analysis

ƒFocusses on:

ƒthe interviewers administration of the questions,


and
ƒthe respondents ability to answer questions
without error

3
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Aims of pre-testing
zIdentify sources of potential non-sample
error:
ƒthe effect the interviewer has on question
responses;
ƒerror the respondent makes due to the
questionnaire;
ƒerrors due to the method of delivery or errors
caused by sequencing; and
ƒerrors caused by the interaction of these
factors.

Benefits of pre-testing
zReduce survey development time
zReduce survey costs
zImprove the quality of data collected

zIdeally all survey development programs


should incorporate some degree of pre-
testing

zPre-testing can minimise non-sampling error


but cannot eliminate all errors in survey data

4
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Types of pre-testing techniques


zLiterature reviews
ƒdatabase and library searches

ƒother statistical agencies and organisations

zExpert reviews
zFocus groups
zCognitive testing
zInterviewer debriefing
zObservational interviews
zBehavior coding

Sampling for pre-tests


zParticipants recruited purposively using
selection criteria specific to the questions to
be tested

z"Questions that fail in the laboratory will


certainly fail in the field but questions that
work in the laboratory may not always work
in the field"

5
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Expert reviews: overview


zGroup of "experts" review questionnaire to
identify potential problems and gather an
understanding of the respondent's task
zList potential problems and classify them to
specific points where difficulties may occur
zWho should conduct an expert review?
ƒAt the ABS, experts means survey design
experts, not subject matter experts, but can
include interviewers

Focus group: overview


zModerator led, informal small group
discussion
zModerator remains neutral at all times
zSeveral groups may be necessary to
represent different population subgroups
zUsed to explore respondent understanding
of, and reaction to terms, concepts and
issues

6
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Cognitive interviews: overview


zCognitive interviews are essentially:
ƒ Asking respondents questions about the
survey questions

zCognitive interviews identify problems in


ƒquestion comprehension

ƒmemory recall

ƒselecting responses

ƒinterpretation of reference periods

ƒreactions to sensitive questions

Cognitive interviews
zConducted one to one with an interviewer in
a standardized environment, usually taped

zCognitive interviews can use


ƒprobing questions

ƒthink alouds

ƒparaphrasing

ƒvignettes

ƒcard sorting tasks

7
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Cognitive interviews
zRetrospective probes
ƒInterviewer administers the survey in totality and

then asks specific probes about particular questions.


ƒUsed to provide information about the survey as a

whole eg, flow and context effects and mode


effects.
zConcurrent probes
ƒRespondents are either asked to verbalise their

thought processes as they answer or they are asked


additional questions straight after a survey
questions.
ƒUsed to understand detailed response processes.

Cognitive interviews
zThink-alouds
ƒThe interviewer asks the respondent to describe

their thought process aloud.


ƒUsed in either concurrent or retrospective

probing techniques.
ƒEG: "Please tell me what you are thinking
about as you answer the next question."

8
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Cognitive interviews
zParaphrasing
ƒThe respondent is asked the question and then is

immediately asked to repeat the question back


in their own words (before they have a chance
to answer the question itself)

Cognitive interviews: overview


zVignettes
ƒParticipant responds to a question from the

point of view of a hypothetical situation.


ƒThis allows interviewers to explore response

processes, especially information about storage


and retrieval of information, regardless of
whether the respondent has experienced the
situation being described, or not.

9
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Cognitive interviews: overview


zCard sorting tasks
ƒProvide interviewers with information about

how respondents think about categories, group


information or define particular concepts.
ƒHighlights the ways in which respondents

define concepts (in terms of inclusions and


exclusions).

Is cognitive interviewing the


best method?
zLeast reliable method for indicating problems
with a questionnaire (Presser & Blair, 1994)

BUT

zThe most detailed pre-testing method for


ensuring respondents are answering questions
as intended
zTechnique that can identify the most detailed
problems with comprehension, retrieval and
judgment processes of the respondent.

10
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Analyzing the data


zWhat attributes of cognitive interviews
make results difficult to generalise?
ƒrespondents are self selecting
ƒpurposive sampling

ƒsmall sample size

ƒnon standardised interview creates between

interviewer variation
ƒlaboratory environment is different to normal

interview environment

Observational interviews
zA trained interviewer observes the survey
process
ƒ eg: form completion, responses during an interview
zCan follow-up with in-depth interview
zIdentifies where respondents seem to stumble
or have trouble
zIdentifies problems in wording, question order,
presentation or layout
zUsually used in testing Self Administered
Questionnaires such as Census forms

11
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Behavior coding
zSystematic coding of respondent and
interviewer behaviors during the interview
zEvaluates the question-asking process and
the administration of the survey by finding
problems in question wording or
understanding
zUses a third party "coder" and a behavior
coding sheet
zCan be done concurrently or retrospectively

Selecting a pre-testing strategy


zNeed right combination of techniques to
achieve the objectives of testing, within
available resources
zFactors to consider
ƒCost
ƒSkilled resources

ƒTimeliness of results

ƒStage of development process

ƒAims of test

ƒTypes of errors to be identified

12
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Selected attributes of pre-


testing techniques
Technique Resources Timeliness of Stage of
results development
process
Cognitive High Fast Middle
Interviews
Focus Groups Medium Fast Start/early
Interviewer Medium Fast Start/early
Debriefing
Expert Review Low Fast Early
Literature Low Moderate Start
Review
Behaviour High Slow Later
Coding
Observational High Moderate Middle
Interviews

Comparison of pre-testing techniques


by sources of error able to be examined
Sources of
error
Focus Interviewer Cognitive
Groups Debriefing Interviews
Interviewer Consistency detailed

Sensitivity detailed
Respondent Comprehens detailed detailed
ion
Recall detailed detailed
Judgement detailed detailed
Response detailed detailed
Burden detailed limited detailed
Sensitivity detailed some detailed

13
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Field Testing
zTwo broad types
ƒPilot Tests

ƒDress Rehearsals

zEvaluation
ƒconducted after each test

ƒconducted after final survey

ƒbased on qualitative and quantitative data

ƒused as a basis for next development cycle

Field Testing (continued)


zSmall operational tests conducted in the
same mode as the final survey

zABS uses:
ƒSample size of about 300 households

ƒSample usually selected randomly

ƒ8 to 12 interviewers involved in tests

14
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Pilot Testing
zPilot testing can provide information on:
ƒidentification of populations

ƒunderstanding of questions (broad)

ƒquestion sequencing and overall survey flow

ƒinterviewer instructions

ƒinterviewer training manuals and procedures

ƒrespondent reaction to questions

ƒediting procedures

ƒprocessing systems (if pilot test data processed)

Dress Rehearsal
zDress rehearsal is the final test and can
provide information on:
ƒall of the items covered in the pilot test
ƒinterview times (important for survey cost)

ƒfield procedure

15
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Changes after tests


zChanges to instruments and procedures can
be made after pilot tests and before the dress
rehearsal

zThe dress rehearsal should be a test of all


elements of the survey in the fashion they
will be conducted in the final survey. There
is only very limited scope to change things
after the DR.

Interviewer debriefing
zDebriefings are:
ƒSubjective and anecdotal

zSurvey interviewers provide feedback on a


previously used survey instrument or draft
questionnaire
zInterviewers are given questionnaires or
participate in focus group style interviews
zAsked to report on any problems they had
administering the questionnaire, using the
assisting documentation, or any particular
reactions or problems respondents had.

16
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Evaluation

zPost survey review


ƒReview all aspects of the survey

ƒUse interviewer feedback on issues relating to the

questions, response categories, procedures etc.


ƒSurvey processing staff feedback on issues of

processing data items, validation etc.


ƒKey user feedback on whether survey data met

their information needs. Identify:


ƒGaps
ƒDeficiencies in collected items
zData quality analyses
ƒNon-response evaluation

Testing reliability/validity of
questions
zOne way to evaluate the validity and reliability of
survey questions is to conduct a follow-up study of
the same respondents and compare survey results
with "gold standard" statistics.
zThis method has been used extensively in testing
Australian census questions on disability to:
ƒ Compare different census test questions
ƒ Conclude that general disability census measure was

not identifying the same population as the survey


measure
zResults very promising for the need for assistance
questions being tested for 2006

17
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Gold standard vs test measure


statistic
z Compare the results of the gold standard and
the test measures (as per the following table)
Test measure
True status + - Total
(Has a disability) (Does not have a
disability)

D a b a+b
(Has a
disability)
not D c d c+d
(Does not
have a
disability)
Total a+c b+d a+b+c+d

Sensitivity and Specificity


zSensitivity is the probability of correctly
identifying someone with a disability (a/(a+b))
zSpecificity is the probability of correctly
identifying someone that does not have a
disability (d/(c+d))
zFalse positive rate is the probability of
identifying someone as having a disability
when they do not (c/(c+d))
zFalse negative rate is the probability of
identifying someone with a disability as not
having a disability (b/(a+b))

18
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Interviewer training
zABS disability survey training sessions were
three days for all interviewers
zInterviewers already well trained in general
interviewing procedures, including computer
assisted interviewing

Interviewer training
(continued)
zTraining covered:
ƒmajor users and uses

ƒstructure of the questionnaire

ƒmain survey populations

ƒquestions and sequencing

ƒcoding of conditions

ƒpractice interviewing

ƒissues relating to interviewing people with

disabilities

19
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Interviewing people with


disabilities
zUnderstand that there are myths and stereotypes
about people with disabilities, but many of them
are not true!!
zAlways interview in a professional and sensitive
manner
zLanguage is important
ƒNeeds to be acceptable to people with disabilities,
for example, neutral and avoiding terms like
disability, handicap, impairment etc.
ƒNeeds to be accurate and respectful of individuals

with disabilities
ƒExamples

Interviewing people with


disabilities (continued)
zAsk all questions
zAsk questions exactly as worded
zDo not assume or anticipate answers
zBe patient and let the respondent finish
zAsk all questions in a positive manner
zMaintain a professional attitude

20
Second Workshop for Improving Disability Statistics and Measurement
Bangkok, 27-29 September 2004

Cultural issues
zThis is a very difficult area
zBeing considered by the WCG on disability
statistics
zSometimes there is no overall word for
"disability"
zReluctance in some cultures to identify people in
their household with a disability
zEncourage identification of people with
disabilities through use of techniques to explain
the purpose of the survey to respondents, such as
primary approach letters and brochures

21

You might also like