(1) Workshop/Seminar on 'The Collection and Analysis of Network Data'
Title: Networks and Behavior: Statistical Models and Advances in the Theory of Action
Venue and date: University of Groningen , The Netherlands, 7-15 September 2006
Senior Instructors:
Tom A.B. Snijders, University of Groningen , The Netherlands
Emmanuel Lazega, Université de Paris IX - Dauphine, France
Course Assistant:
Ruud van der Horst, University of Groningen, The Netherlands
Workshop Summary:
This workshop presents recent advances in statistical models for social networks and their substantive application in social science, generally aiming at the interdependence between networks and behavior. Behavior is understood to cover not only concrete action but also attitudes, performance, etc. It has been widely recognized that the network embeddedness of social actors influences their behavior, and that their behavior in its turn influences their choices of interaction partners and thus the network, but statistical methods to study this mutual influence have only recently started to be developed.
Models will be treated both for non-longitudinal and for longitudinal network data, but the latter shall receive more attention as they allow a richer study of the mutual influences between how social actors are embedded in networks, and their behavior. The workshop covers both the statistical methodology, the use of the SIENA software to apply these methods, and substantive applications. The applications treated are in the field of organization studies, but depending on the interests of participants, other applications can also be discussed (in more limited ways, of course). Network data analysis will be used for examining the key social processes of solidarity and learning, as well as one or more of the following: control, power, brokerage, regulation. Methodological and substantive sessions will alternate to some extent, but the start of the workshop will focus on methodology while the later days will focus on substantive applications.
Statistical Methods for Studying Social Networks:
The methods studied are about complete networks, i.e., data sets about a given group of social actors, in which it is known which ties exist, ties being defined according to one or more well-defined relational criteria (e.g., collaboration, esteem, friendship, exchange, etc.). The strong dependencies between the variables indicating such ties lead to complications in the statistical model.
For non-longitudinal data, the p* or ERGM (Exponential Random Graph Model) has been developed. Difficulties in estimating and applying this model have recently been alleviated by the development of new model specifications; see Snijders, Pattison, Robins, and Handcock (2006). A methodology that is perhaps eventually more scientifically useful, although it poses the stronger requirement that longitudinal (panel) data must be available, is the actor-oriented approach to network dynamics (Snijders, 2001, 2005). This approach is currently being extended to statistical models for analyzing panel data on networks and behavior simultaneously (Steglich, Snijders, and Pearson, 2004; Snijders, Steglich, and Schweinberger, 2006). The actor-oriented approach is based on the assumption that actors have an active role in defining and changing their network as well as their behavior, and yields a statistical model that agrees well with substantive social science theories.
The statistical methods are implemented in the SIENA program (available from http://stat.gamma.rug.nl/stocnet/), which will be available for the participants in the workshop and which will be the basis for the practical work.
Substantive Material:
Substantive applications of network analysis must be framed by a theory
of individual and collective action. Any theory of collective action makes behavioral
assumptions about actors' rationality and their capacity to negotiate and renegotiate
their participation and commitment. In order for such a theory of action to
be able to use network analysis, it must allow for a view of actors not as short-sighted
and disconnected utility maximizers, but as individuals capable of social exchange
of multiple resources, of commitment to exchange partners, of complex interactions,
and of relational signalling. Social and economic networks are then seen as
indicators of resource interdependencies among these actors. In this workshop,
a broadly conceived and multilevel rational choice framework will be used, in
which social actors can be regarded as being characterized by a contextual
and strategic rationality.
Case studies used in the workshop will be based on organization research. In
such contexts, individual members' formal positions and property rights are
not enough to guarantee the functioning of the organization as a collective
actor. For example, to do their work, individual members use social relations
to access and exchange resources on an ongoing basis. They invest in relationships
and these investments trigger specific social processes that produce certain
forms of public good (public within the organization). Among these goods, on
can include various forms of bounded solidarity (measured by direct and indirect
reciprocity), an indirect control regime, or a regulation system stabilizing
the redefinition or renegotiation of rules (by bestowing legitimacy on a selection
of unofficial representatives) that takes place when the organization needs
to adapt to exogenous as well as endogenous pressures for change. The organization
works because both an institutional arrangement and social processes based on
resource interdependencies make it worthwhile for members to undertake socially
productive activities. These processes are not the result of a deus ex machina
; they are consistent with individual interests and with members' management
of resource interdependencies and relationships. In other words, they are also
the result of a form of social discipline. Social networks can thus be seen
to be indicators of the existence of such a discipline, and their analyses be
used to identify the characteristics of this discipline.
These social processes can be examined using all the methods of Network analysis
that will be taught during the first days of this Workshop. For example, indicators
for solidarity in an organization can be found by looking for cohesive subgroups
and various forms of blockmodels in a particular network or across several networks
simultaneously; we can also use p2 or p* models to test for the importance of
direct and indirect reciprocity in such networks. Another example is the study
of informal social control in organizations: combining sociometric data with
information on manipulation of relationships by members of a collective who
react to opportunistic or deviant behaviour by using their interdependencies
and by trying to pressure each other back to good order. The political process
of rule definition and influence can also be tracked by a careful analysis of
various centrality measures in these networks, identification of forms of status
inconsistency, correlation between such forms and positions taken in management
policy making, etc. Finally, longitudinal analyses using Siena will be used
to look at the social process of collective learning over time. No a priori
limitation will be set in the use of network methods for the study of social
processes.
This workshop focuses on complete networks, because analytical primacy is given
to the study of social systems (meso-level). The use of (longitudinal information
about) relatively small scale organizational systems is particularly well suited
to illustrate the working of different social mechanisms, since organizations
deliberately define the context conditions for individual action (e.g. in terms
of incentive structures and functional interdependencies). Furthermore, focusing
on organizations facilitates the solution of the boundary problem (because organizations
have a formal criterion of membership).
Data Sets:
I. The law firm dataset comes
from a network study of corporate law partnership that was carried out in the
early 1990's in New England . It includes measurements of networks among the
71 attorneys (partners and associates) of this firm, including their strong-coworker
network, advice network, friendship network, and indirect control networks.
Various members' attributes are also part of the dataset, including seniority,
formal status, office in which they work, gender, lawschool attended, individual
performance measurements (hours worked, fees brought in), attitudes concerning
various management policy options, etc. This dataset was used to identify social
processes such as bounded solidarity, lateral control, quality control, regulation,
etc. among peers. The ethnography, organizational and network analyses of this
case are available in Lazega (2001). The dataset will be made available to the
workshop participants who will each be given a CD, which they can take home
after the workshop.
II. The Commercial Court of Paris dataset ,
an organizational longitudinal network study data set that was collected from
2000-2005 at this courthouse. A total of around 200 "consular" judges
participated in the survey. The panel consisted of three waves with intervals
of between two and three years. On the individual level, the dataset contains
information on a.o. different kinds of decisions elicited through vignettes,
opinions about institutional redesign, etc. Social network measures reconstituted
the advice network among the judges. The dataset also includes information about
advice seeking before recruitment as a judge and about participation in social
events organized by the courthouse. The first two waves of this dataset will
be made available for training purposes during the workshop. Background information
on the data and data collection can be found in the references.
References:
Methodological
- Robins, G., P. Pattison, Y. Kalish, & D. Lusher (2006) An introduction to exponential random graph ( p* ) models for social networks. Social Networks . Forthcoming.
- Robins, G., Snijders, T., Wang, P., Handcock, M., & Pattison, P. (2005). Recent developments in Exponential Random Graph ( p* ) Models for Social Networks. Social Networks . Forthcoming.
- Snijders, Tom A.B., The statistical evaluation of social network dynamics. Pp. 361-395 in Sociological Methodology - 2001 , edited by M.E. Sobel and M.P. Becker. Boston and London : Basil Blackwell.
- Snijders, Tom A.B. (2005). Models for Longitudinal Network Data. Chapter 11 in P. Carrington, J. Scott, & S. Wasserman (Eds.), Models and methods in social network analysis . New York : Cambridge University Press.
- Snijders, Tom A.B., Pattison, Philippa E., Robins, Garry L., and Handcock, Mark S. (2006). New specifications for exponential random graph models. In press, Sociological Methodology .
- Snijders, Tom A.B., Steglich, Christian E.G., and Schweinberger, Michael, Modeling the co-evolution of networks and behavior. To appear in Longitudinal models in the behavioral and related sciences , edited by Kees van Montfort, Han Oud and Albert Satorra; Lawrence Erlbaum, 2006.
- Tom A.B. Snijders, Christian Steglich, Michael Schweinberger and Mark Huisman. 2006. Manual for SIENA , version 2.4. http://stat.gamma.rug.nl/stocnet/
Steglich, C.E.G., Snijders, T.A.B. and Pearson, M. (2004). Dynamic Networks and Behavior: Separating Selection from Influence . Submitted for publication.
Wasserman, S. & P. Pattison. 1996. "Logit models and logistic regressions for social networks: I. An introduction to Markov graphs and p* ." Psychometrika 61, 401 425.
Substantive
- Ana Maria Falconi, Karima Guenfoud, E. Lazega, Claire Lemercier, Lise Mounier
(2005), "Le Contrôle social du monde des affaires: une étude
institutionnelle", L'Année sociologique, 55(2):451-484.
- Lazega E. (2001), The Collegial phenomenon: The social mechanisms of cooperation among peers in a corporate law partnership , Oxford : Oxford University Press.
- Lazega, E., Claire Lemercier, Lise Mounier (2005), "Dynamics of advice networksand intra-organizational learning: A spinning top model".
- Lazega, E. and Lise Mounier (2003), "Interlocking Judges: On Joint (External and Self-) Governance of Markets", in Vincent Buskens, Werner Raub and Chris Snijders (eds), Research in the Sociology of Organizations , 20: 267-296.
Seminar Presentations:
- Acculturation revisited: a model of personal network change, Jose Luis Molina and Miranda Lubbers, Universitat Autonoma de Barcelona, Spain (co-author Chris McCarty),
- A Spinning top model of formal organization and informal bahavior: dynamics of advice networks among judges in a commercial court, Emmanuel Lazega, Université de Paris IX - Dauphine, France
- Structural Equivalence and International Conflict: A Social Networks Analysis, Ilan Talmud, University of Haifa, Israel
- Network Centrality and International Conflict, 1816-2001: Does it Pay to Be Important? Ilan Talmud, University of Haifa, Israel
- Research groups` social capital and PhD students performance: The case of Slovenia, Anuska Ferligoj, University of Ljubljana, Slovenia (co-authors Petra Ziherl and Hajdeja Iglic)
- Why more contact may increase cultural polarization. Extending models of cultural influence in dynamic interaction networks, Andreas Flache, University of Groningen, The Netherlands (co-author Michael W. Macy, Cornell)
- Dynamic Network Analysis of Text, Ulrik Brandes, University of Konstanz , Germany
- Social Context and Network Formation: An Experimental Study, Vincent Buskens University of Utrecht , The Netherlands
- Trust and stress in organizations; testing Krackhardt's "philos" hypothesis, Filip Agneessens, Ghent University, The Netherlands and Rafael Wittek, University of Groningen, The Netherlands
- The network dynamics of mergers and acquisitions in the international electricity industry, Alessandro Lomi, University of Bologna , Italy
- Some basic methodological issues concerning statistical modeling of networks, Tom Snijders, University of Groningen, The Netherlands
Title: Models for Longitudinal and Incomplete Data
Venue and date: Hasselt University (formerly Limburgs Universitair Centrum), Hasselt , Belgium , 13-21 September 2006
Senior Instructors:
Geert Molenberghs , Center for Statistics, Hasselt University , Belgium
Geert Verbeke , Biostatistical Centre, Katholieke Universiteit Leuven , Belgium
Workshop Summary:
Based on Verbeke and Molenberghs (2000), we first present linear mixed models for continuous hierarchical data. The focus lies on the modeler's perspective and on applications. Emphasis will be on model formulation, parameter estimation, and hypothesis testing, as well as on the distinction between the random-effects (hierarchical) model and the implied marginal model. Apart from classical model building strategies, many of which have been implemented in standard statistical software, a number of flexible extensions and additional tools for model diagnosis will be indicated. A number of illustrations and worked examples will be given based on the SAS procedure MIXED.
Then, extensions will be formulated to model outcomes of a categorical nature, including counts and binary data. Based on Molenberghs and Verbeke (2005), several families of models will be discussed and compared, from an interpretational as well as a computational point of view.
To begin with, models will be discussed for the full marginal distribution of the outcome vector. This allows model fitting to be based on maximum likelihood principles, immediately implying inferential tools for all parameters in the models. The main disadvantage of such models is that they require complete specification of all higher-order interactions, which is often based on unrealistic assumptions, and often lead to computational problems, especially in examples with many repeated measurements per subject.
Alternatively, semi-parametric methods can be used which no longer require full specification of the likelihood, only of the first moments or of the first and second moments. This leads to the so-called generalized estimating equations. Estimation and inference will be discussed and illustrated in full detail.
Then, mixed-effects models for non-Gaussian data will be discussed, with a strong emphasis on the generalized linear mixed model. Similarities and differences with the linear mixed model will be discussed. Care will be taken in explaining the numerical difficulties encountered when fitting this type of models to data, since the optimization process involves non-trivial integration. Numerical integration will be contrasted with alternative methods, such as, for example, PQL and MQL. Marginalization of the generalized linear mixed model will be given detailed treatment. In particular, attention is devoted to the fact that parameters in marginal and random-effects models require different interpretation.
The primary software tools for the non-Gaussian part will be the SAS procedures GENMOD, GLIMMIX, and NLMIXED, but a number of other tools will be discussed as well.
Finally, when analyzing hierarchical and longitudinal data, one is often confronted with missing observations, i.e., scheduled measurements have not been made, due to a variety of (known or unknown) reasons. It will be shown that, if no appropriate measures are taken, missing data can cause seriously jeopardize results, and interpretational difficulties are bound to occur. Methods to properly analyze incomplete data, under flexible assumptions, are presented. Key concepts of sensitivity analysis are introduced.
Classroom lectures alternate with hands-on practical sessions.
Method of working:
As a result of the course, participants should be able to perform a basic analysis for a particular longitudinal data set at hand, using linear and generalized linear models for longitudinal data. Based on a selection of exploratory tools, the nature of the data, and the research questions to be answered in the analyses, they should be able to construct an appropriate statistical model, to fit the model within the SAS framework, and to interpret the obtained results. Further, participants should be aware not only of the possibilities and strengths of a particular selected approach, but also of its drawbacks in comparison to other methods.
The course will be explanatory rather than mathematically rigorous. Emphasis is on giving sufficient detail in order for participants to have a general overview of frequently used approaches, with their advantages and disadvantages, while giving reference to other sources where more detailed information is available. Also, it will be explained in detail how the different approaches can be implemented in the SAS package, and how the resulting outputs should be interpreted.
Datasets:
Throughout the classroom lectures, a variety of real data applications will be used to motivate and illustrate methodology. These come from the two textbooks listed below. Consequently, many are of a biomedical nature. In the hands-on sessions, selected applications from social sciences and related areas will be used. This will allow participants to see how the same methodology and methods of analysis can be used throughout a variety of areas. Finally, course participants can bring their own datasets for discussion and analysis during the practical sessions. To the extent possible, such datasets will be used during the practical sessions.
Pre-requisites:
Throughout the course, it will be assumed that the participants are familiar with basic statistical modeling, including linear models (regression and analysis of variance), as well as generalized linear models (logistic and Poisson regression). Moreover, pre-requisite knowledge should also include general estimation and testing theory (maximum likelihood, likelihood ratio). Some familiarity with the SAS system for linear and generalized linear models will be of great benefit to the participants. Those who have not used SAS before are recommended to familiarize themselves with the system prior to the workshop. Both books references provided in the literature section contain chapters and section dedicated to the use of the SAS system. Additional guidance will be provided at the workshop for such new users of SAS.
Literature (books, readers):
• Copies of the transparencies used in the course.
• Verbeke, G. and Molenberghs, G. (2000) Linear Mixed Models for Longitudinal Data . New York : Springer-Verlag.
• Molenberghs, G. and Verbeke, G. (2005) Models for Discrete Longitudinal Data . New York : Springer-Verlag.
Seminar Presentations:
- A measure for the reliability of a rating scale based on longitudinal clinical trial data, Annouschka Laenen, University of Hasselt, Belgium
- The impact of a misspecified random-effects distribution on maximum likelihood estimation in generalized linear mixed models, Saskia Litière and Ariel Alonso, University of Hasselt, Belgium
- Sensitivity analysis for shared parameter models using copulas, Dimitris Rizopoulos, Catholic University of Leuven, Belgium
- Random effects for multivariate repeated responses, Geert Verbeke, Catholic University of Leuven, Belgium
- Multivariate one-sided tests in constrained parameter spaces, Geert Molenberghs, University of Hasselt, Belgium
- Shared parameter models with a flexible random effects distribution, Spyridoula Tsonaka, Catholic University of Leuven, Belgium
- A simulation study comparing weighted estimating equations with multiple imputation based estimating equations, Cristina Sotto, University of Hasselt , Belgium
- A hierarchical Ornstein-Uhlenbeck process for core affect trajectories, Francis Tuerlinckx, Catholic University of Leuven, Belgium
- Kernel weighted influence measures, Niel Hens, University of Hasselt, Belgium
- Characterizing persistent disturbing behaviour using longitudinal and multivariate techniques, Jan Serroyen, University of Hasselt , Belgium
- Sensitivity analysis for longitudinal clinical trials, Herbert Thijs, University of Hasselt , Belgium
Title: Theory-Driven Evaluation and Intervention Studies in the Social Sciences
Venue and date: University of Cyprus, Nicosia, Cyprus, 21-29 September 2006
Senior Instructors:
Peter Schmidt, University of Giessen , Germany
Sebastian Bamberg, University of Dresden , Germany
Eldad Davidov, University of Basel , Switzerland
Workshop Summary:
The course is going to systematically cover our research. We are planning to discuss in the course the following topics:
1) Providing a guideline for evaluating intervention studies.
2) Combining theoretical as well as methodological aspects of policy related evaluation in different policy fields. As an example we are going to discuss the topic of intervention studies in ecological behavior in general, and travel mode choice in particular. Furthermore, we will deal with intervention studies in the field of labour market research.
3) On the theoretical level we are planning to work intensively on the concept of theory driven evaluation (e.g. Chen, 1990) that is the systematic application of social science action theories for modelling the social / psychological factors causally mediating people's reaction to an intervention aiming to change their behavior.
4) We will discuss and apply another interesting approach in this field, the development of the so called 'logical models' (e.g. Weiss, 1995) as the more qualitatively oriented approach of 'realistic evaluation' (Tilly et al. 1997).
5) On the methodological level we are planning to practically apply classical and new evaluation designs (e.g. group randomized trials).
6) We are going to teach new statistical methods for analysing the behavioral effects of interventions (e.g. latent growth curve models, autoregressive cross-lagged models, a synthesis of autoregressive and latent growth curve models).
Since more than 10 years, our research group in Giessen on "Changing processes" financed by the German Science Foundation has been conducting research on the evaluation of transport policy measures. We have conducted a lot of large scale experimental and quasi-experimental field-studies evaluating the effect of measures to motivate car users to use more frequently non-motorized travel means. Thus we offer to use our research as a starting point for a strongly practice-oriented evaluation course. This course may contain an introduction in evaluation designs, theoretical evaluation frameworks and data analysis. In sum, we would like to train participants ability to apply theoretically as well as methodologically sound evaluation designs for their own research. For this purpose we would offer participants to work on case studies and reanalyse our data sets. Alternatively, participants may bring their own data, and analyse it with tools provided in the course.
General Literature
-Chen, H.-T. (1990). Theory Driven Evaluations, Newbury Park : Sage.
-Rossi, P.H. and H.E. Freeman (1993). Evaluation. A systematic approach, fifth edition, Newbury Park , London , New Delhi : Sage.
-Shadish, WR, Cook, TD, & Campbell, DT (2002). Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston : Houghton-Mifflin.
-Bollen, K. and P. Curran (2006, in press). Latent Curve Models. New York : Willey.
Publications of the working group on theory-driven evaluation
-Bamberg , S. & Schmidt, P. (1998). Changing Travel-Mode Choice as Rational Choice: Results from a Longitudinal Intervention Study. Rationality and Society, 2, 223-252.
-Bamberg , S. & Schmidt, P.(1999) Regulating transport: Behavioral changes in the field. Journal of Consumer Policy, 22, 479-509.
-Bamberg , S. (2000) The Promotion of New Behavior by Forming an Implementation Intention - Results from a Field-Experiment. Journal of Applied Social Psychology,30,9, 1903-1922.
-Bamberg , S. & Schmidt, P. (2001). Theory-driven Evaluation of an Intervention to reduce the private Car-Use. Journal of Applied Social Psychology, 31,6, 1300-1329.
-Bamberg , S. (2002). Implementation Intention versus Monetary Incentive. Comparing the Effects of two Interventions to promote the Purchase of Organically Produced Food. Journal of Economic Psychology, 23(5), 573-87.
-Bamberg , S. (2002). Effects of Implementation Intentions on the actual Performance of New Environmentally Friendly Behaviours - Results of two Field-Experiments. Journal of Environmental Psychology, 22, 399-411.
-Bamberg , S., Rölle, D., Weber, C. (2003). Does habitual car use not lead to more resistance to change of travel mode? Transportation, 30, 97-108.
-Bamberg , S. & Schmidt, P. (1996). Theory-driven evaluation of environmental policy measurements A regional case-study. In: R.K.aufmann-Hayoz. Bedingungen umweltverantwortlichen Handelns von Individuen. Proceedings of the Interdisciplinary Symposium "Responsible Environmental Behavior", Berne , September 1996. Interfakultäre Koordinationsstelle für Allgemeine Ökologie, Universität Bern .
-Bamberg , S. & Schmidt, P. (1998). Modeling the Dynamics of Micro-Social Change: Results of a Three Wave Intervention of Travel-Mode Choice in a Region. In: H.-P. Blossfeld & G. Prein. Rational Choice Theory and Large-Scale Data Analysis. pp. 258-278. Boulder/Oxford: Westview Press.
-Davidov, E., Schmidt, P. and Bamberg , S. (2004). Modeling Longitudinal Data of an Intervention Study on Travel Model Choice: Combining Latent Growth Curves and Autoregressive Models. In van Montfort, K., Oud, H. and Satorra, A. (Eds), Recent developments in structural equation modeling: theory and application. The Netherlands : Kluwer Academic Publishers.
-Rölle, D., Weber, C., & Bamberg , S. (2002). Modified Mobility Behavior at the new residence? - The role of information in changing the choice of transportation moving to a new city . In: M. Möhlenbrink, M. Bargende, U. Hangleiter, U. Martin (Eds.), Networks for Mobility. Proceedings of the international symposium 2002, Stuttgart . Stuttgart : University of Stuttgart , Centre of Transportation Research.
-Yang-Wallentin, F., Schmidt, P., Davidov, E. and Bamberg , S. (2004). Is there any interaction effect between intention and perceived behavioural control? Methods of Psychological Research Online, 8(2), 127-157.
-Yang-Wallentin, F., Schmidt, P. & Bamberg, S. (2001). Testing interactions with three different methods in the theory of planned behavior: analysis of traffic behavior data. In R. Cudeck et al. (Eds.), Structural Equation Models: Present and Future. (405-423).Lincolnwood, IL: Scientific Software International, Inc.
Seminar Presentations:
- The Analysis of Individual and Average Causal Effects: Basic Principles and Some Applications, Rolf Steyer, University of Jena, Germany
- Realist Evaluation and Realist Synthesis, Raymond Pawson, University of Leeds, United Kingdom
(4) Workshop/Seminar on 'Measurement, Data Collection and Data Quality Issues for Cross-National Survey Data'
Title: Measurement, Data Collection and Data Quality Issues for Cross-National Survey Data
Venue and date: Università della Svizzera Italiana (USI), Lugano, Switzerland, 17-25 August 2006
Senior Instructors:
Willem E. Saris, Faculty of Social and Behavioral Science, University of Amsterdam, The Netherlands
William van der Veld, Amsterdam School of Communication and Research, University of Amsterdam, The Netherlands
Course Assistant:
Daniel Oberski, University of Amsterdam, The Netherlands
Workshop Summary:
In Europe several large scale cross-national surveys are done. The most well known are The Eurobarometers, European Social Survey, European Value study and the European Election study. Because of the difference in language between the different countries an extra problem exists in cross-national surveys compared with national survey.
In general one can expect systematic and random errors in survey research. They can cause bias in response distributions, means and other summary statistics and larger standard errors of these estimates. These errors can also lead to biased estimates of the relationships between variables and large standard errors of correlation and regression coefficients and other measures for the relationships between variables.
The most important sources of errors in national surveys are sampling designs, non-response and measurement error. Sampling error and non-response cause mainly bias in response distributions and summary statistics while measurement error can affect these characteristics but can also seriously bias measures for relationships. In Europe with its different countries and languages a special problem is that these errors can make the results of studies in the different countries incomparable.
Given this situation, it is important for researchers to know how to detect such errors and how to correct for them. In the course we concentrate on measurement errors but in the seminar at the end also the other issues will be introduced. In the workshop we teach people on the basis of data of the European Social Survey (ESS) how these errors can be detected and corrected so that the results for the different countries can be compared.
The general idea of the course:
In operationalisation and data collection more than 50 decisions have to be made in developing a single question. For example one has to decide about: open and closed questions, number of categories, use of batteries with statements or questions etc. These decisions can have considerable effects on the response distributions as has been shown in many studies (Schuman and Presser 1981, Tourangeau et al. 2000) but also on the relationships between variables (Andrews 1984, Scherpenzeel and Saris (1997, Saris, Van der Veld and Gallhofer 2004). Also cross-nationally considerable differences can be obtained due to differences in methods in different countries (Van der Vijver and Harkness 2004). Therefore it is impossible to compare results across countries without correction for measurement error.
As was said before one can make a distinction between systematic errors (method effects causing invalidity) and random errors (causing lack of reliability). If one knows how these decisions have an influence on the data quality i.e. the reliability and the validity, one can also correct the biases in the estimates of relationships.
The objective of the course is to prepare the students to conduct a specific task from the beginning to the end of the Workshop; ideally this goal will be successfully reached if they leave the Summer School with a draft of an empirical research paper that shows theoretical and methodological understanding of the problems of European comparative research. It means that the teaching model is one in which theory and methodology of empirical data analysis is properly mixed and students will have ample possibilities for data analysis themselves.
For the practical work in the afternoons we will use the ESS data on social capital variables: social contacts, social trust and political trust. For this topic data of two rounds are available from more than 22 countries while in the supplementary questionnaire alternative measures for the same variables are obtained. Use of multiple methods allows seeing the effects on the results of differences in methods within and across countries.
Preparatory Reading:
We suppose that all participants are familiar with regression analysis, survey design and analysis and the use of SPSS.
In order to avoid teaching of material one can easily read individually, we suggest that the participants read in advance the texts mentioned below which will be sent to them before the course starts.
- Chapters 1-8 of the textbook of W.E. Saris and I.N. Gallhofer: Design, evaluation and analysis of questionnaires for survey research. The rest of the book will be presented and discussed in the course.
- Pippa Norris: Civic Society and Social Capital
- Ken Newton: Social Trust and political disaffection, social capital and democracy.
We expect the participants to take a look at the questionnaires. Especially they should look at the measurement of social contacts, social trust and political trust in the ESS questionnaire of round 1 and 2 as well as the questions in the supplementary questionnaires where alternative forms for the same measures are presented to the respondents.
Please click here to see Norris and Newton articles and ESS questionnaire.
Seminar Presentations:
- Bringing Values Back In: A Multiple Group Comparison with 20 Countries Using the ESS 2003. Measurement, Causes and Consequences, Eldad Davidov, University of Basel, Switzerland
- Assessing Cross-National Construct Equivalence in the ESS: the Case of Religious Involment, Jaak Billiet, Catholic University of Leuven, Belgium
- Models for Multiple Group and Heterogeneous Data, Albert Satorra, Universitat Pompeu Fabra, Spain
- Assessing Cross-National Construct Equivalence in the ESS: the Case of Six Immigration Items, Jaak Billiet, Catholic University of Leuven, Belgium
- Effective Sample Size. Theoretic Perspectives and Empirical Insights from the ESS, Mathias Ganninger, Research Centre for Survey Research and Methodology, Germany
- Estimation of Response bias in the ESS: Using Information from Reluctant Respondents in Round One, Jaak Billiet, Catholic University of Leuven, Belgium
- Can we improve the response? Experiments in Switzerland, Dominique Joye, University of Neuchâtel, Switzerland
- Nonresponse in Survey Research: Why is It a Problem? Robert Voogt, Dutch Ministery Of Social Affairs and Employment, The Netherlands