9th ESS Train: At the ECPR Summer School

Course C: "Accuracy of survey estimates: sampling, weighting, variance estimation and design effects in cross-sectional surveys"

7-8 August 2010, Ljubljana, Slovenia

Instructors: Professor Ralf Münnich, Dr. Matthias Ganninger

Many of the participating countries of multi-national sample survey projects like the European Social Survey (ESS) or the International Social Survey Project (ISSP) apply complex sample designs (e.g. cluster-, multi-stage or stratified sample designs). These sample designs have their merits and drawbacks which must be taken into account in the data analysis process.

Today, data analysts can choose from a wide range of data analysis techniques (e.g. design weighting and calibration, weighted least squares regression, etc.) which consider the specifics of a given sample design. The informed choice of a specific technique, however, requires detailed knowledge about its appropriateness for the data to be analyzed.

Nowadays, accuracy measurement for given statistics becomes more and more important. In general, accuracy measurement is based on the use of variance estimation methods. Although some modern statistical software packages like SAS, STATA and R provide users with a (limited) range of variance estimation methods, the decision for or against an adequate technique is left to the data analyst. A thorough knowledge of the methods’ functioning should guide this decision as well as possible peculiarities in the data. Furthermore, with complex sample designs the emergence of design effects can have a negative influence on the precision of estimators. Treating the sample data naively as having arisen from a simple random sample can lead to an underestimation of the variance of an estimator. In comparative studies, the design effect shall give additional information to simply using variance estimation. Different approaches to the estimation of design effects exist from which different estimators can be derived. Which estimator to choose depends on the structure of the data at hand.

The objective of this lecture is to provide the participants with an overview of the recent findings on the use of sample designs, weighting and variance estimation techniques in data analysis and to develop research ideas on how to push the field even further. The course will help participants gain an understanding of the impact of the sample design on estimators and show pathways to an increased quality of their research results. References and a power-point handout will be provided to all participants.

Prof. Dr. Ralf Münnich, Economics and Social Statistics Department at the University of Trier, Germany.

Ralf Münnich is full professor at the University of Trier. He obtained his PhD in survey statistics at the University of Tübingen (1996) and received the venia legendi in statistics and operations research in 2005. His main activities focus on small area statistics applications and variance estimation methods with recent applications to census methods and poverty measurement. Münnich was responsible for several large project like DACSEIS, KEI, and AMELI (EU projects) as well as for the German Census sampling project. Münnich is associate editor of Advances in Statistical Analysis, Wirtschafts- und Sozialstatistisches Archiv, Survey Research Methods, and International Journal of Statistics and Management System. He has published many articles in national and international scientific journals. His most recent publications focus on small area statistics and its applications as well as on variance estimation methods.

He has taught specialized courses on survey statistics and variance estimation methods in PhD courses. More information is available at http://www.uni-trier.de/index.php?id=4761

Dr. Matthias Ganninger, Researcher, GESIS – Leibniz Institute for the Social Sciences, Mannheim, Germany.

Matthias Ganninger has received a doctorate for a thesis on design effects at the University of Trier and published many articles in international scientific journals. He co-authored a textbook on data analysis in the social sciences. He has taught specialized courses on survey methodology in Germany and Europe. More information is available at http://www.gesis.org/das-institut/mitarbeiter-adressen/mitarbeiterverzeichnis/?valpha=&selcat=G%3E%3E&order=sortname&id=145&&selres=63&pagecount=1#63