Quality control

• Pages: 25-30

Quality control

Benefit s of gender equality th rough infrastruct ure provision: an EU-wid e survey 25

6. Quality control

Data quality was ensured during the whole data

life cycle, across the processes of planning,

implementation and assessment.

During the planning stage, the methodology

(e.g. items of the questionnaire, number, loca -

tion and timing of the samples to be collec ted)

was key to ensuring data quality. The concepts

under study used measurements tested in ear-

lier studies. As the questionnaire needed to be

made available in 25 dierent languages by the

contractor of the EU Member States, pre-tests

were carried out to ascer tain whether ques-

tions and answer scales were clear, complete

and correct.

During the implementation phase, data-collec-

tion specications were followed to ensure that

the data collected were sound, robust and of

the highest quality. Quality assurance on com-

parability required clear denition and state-

ment instructions to be followed.

• Coverage of sample, representing the target

population regionally and across types and

size and comparable across Member States.

• Non-response computed according to one

internationally accepted standard across all

participating Member States.

• Non-response bias estimated in a stand-

ard fashion across all participating Member

States.

• List of sampling management sys tems used;

indication of risk due to possible dierences

in eldwork across Member States due to

use of dierent sampling management sys-

tems.

• List of CATI systems used; indication of risk in

using dierent C ATI systems.

• List of the interview sta composition (age,

gender, education) for each Member State;

indication of risk due to dierent sta com-

position.

• Documentation of translation process and

outcomes; indication of risk involved due to

translation or adaptation.

Data checks were carried out on the pre-test

data then on the soft launch. Completed inter-

views were checked during and af ter eldwork.

Phone les of the population under study had

to meet established quality standards, includ-

ing checking for missing or erroneous data, l-

tering, interview duration, duplicates, straight

liners, cross-consistency, outliers and non-re-

sponses.

The data-collection system was computerised.

Fieldwork and sample quot as were closely mon-

itored throughout, with regular data check s

during the eldwork.

Interviewers at tended an in- depth brieng.

Supervisors then conducted qualit y control dur-

ing the eldwork by listening to inter views and

providing the inter viewers with feedback. Thor-

ough data cleaning took place after the eld-

work.

An important indicator of the survey quality is

the eec tive RR: higher RRs ensure more accu-

racy in the survey. The RR was satisfactory in

most cases using the t wo estimation proce-

dures (see Section 5 for detail).

During the assessment stage, data were vali-

dated through an in-depth statistical analysis,

focusing on the following aspects.

• Error detection and debugging data le.

• Analysis of missing values.

• Data ltering.

• Analysis of reliability, validity and robustness.