enEnglish
CROS

Non-response problems and data imputation methods

 

Non-response problems and data imputation methods

Course Leader

Barry SCHOUTEN

Target Group

All NSIs staff dealing with data collection facing non-response, either unit non-response where entire units intended to be collected are missing or item non-response where some items of otherwise responding units are missing

Entry Qualifications

  • Sound command of English. Participants should be able to make short interventions and to actively participate in discussions
  • An academic degree with some knowledge of statistics (social science, statistics, economics), especially multivariate statistics
  • Some basic knowledge of survey sampling and statistical modelling

Objective(s)

  • The main objective of the courses is to enhance the theoretical and practical knowledge related to the treatment of unit non-response and item non-response. In particular, participants will gain knowledge on weighting techniques in order to deal with unit non-response and imputation techniques in order to deal with item non-response. For unit-nonresponse, participants will also learn about up to date monitoring of data collection and application of adaptive survey designs.

Contents

  • The course covers both person and household surveys and business surveys;
  • The non-response problem. Definition of non-response, causes of non-response, types of non-response (unit, item), calculation of response rates;
  • Some mathematical background, models for non-response, effects of non-response on bias, confidence intervals, MCAR (Missing Completely at Random), MAR (Missing at Random), NMAR (Not Missing at Random);
  • How to detect non-response bias, role of auxiliary information. Selection of important auxiliary variables;
  • Monitoring of unit-nonresponse and adaptive survey designs;
  • Correcting afterwards for unit non-response, weighting adjustment techniques (post-stratification, linear weighting, multiplicative weighting, calibration, propensity score method), central-question-approach, re-sampling non-respondents;
  • Correcting afterwards for item non-response: mean/mode imputation, (hot-deck and cold-deck) donor imputation, regression imputation, predictive mean matching, longitudinal imputation (panel data), last information carry forward;
  • Communication to different user groups on the level of non-response, on the correction methods applied and on the impact on accuracy of results.
  • Integration of the treatment of unit- and item-nonresponse methods
  • Gaining experience through realistic, practical computer exercises.

Expected Outcome

Good understanding of the non-response problem. Knowledge of correction methods for unit and item non-response, and the ability to apply these methods in practice.

Training Methods

  • Lectures about non-response problems and correcting methods for unit and item non-response
  • Some practical cases
  • Practical exercises based on real data sets
  • Reading material (course books + additional papers)
 

Required Reading

None

Suggested Reading

  • Jelke Bethlehem, Fannie Cobben & Barry Schouten (2011), Handbook of Non-response in Household Surveys. Wiley & Sons, Hoboken, NJ
  • Little, R.J.A. and D.B. Rubin (2002), Statistical Analysis with Missing Data (second edition). John Wiley & Sons, New York.
  • Schouten, B., Peytchev, A., Wagner, J. (2017), Adaptive Survey Design, Chapmann & Hall

Required Preparation

Participants are invited to sketch non-response in their own organisation, both unit non-response as well as item non-response: magnitude, treatment, etc.

Participants are also invited to present their own survey case studies and list questions they may have. These will receive extra attention in the course.

Trainer(s)/
Lecturer(s)

Prof. dr. Barry Schouten

Prof. dr. Ton de Waal

Practical Information

When

Duration

Where

Organiser

Application  via National Contact Point

21-22.01.2021

2 days

ONLINE

ICON- INSTITUT Public Sector GmbH

Deadline: 10.12.2020