Von 14. bis zum 16. April 2010 findet im 
Zentrum für interdisziplinäre Forschung (ZiF) ein vom Zentrum für Statistik veranstalteter Workshop unter der Leitung von Göran Kauermann (Bielefeld), Harry Haupt (Bielefeld), Jost Reinecke (Bielefeld) und Mark Stemmler (Bielefeld) zum Thema "Statistical Analysis of Longitudinal/Panel Data Structures Arising from Complex Interdisciplinary Questions" statt.
Zum Workshop ist ein Special Issue in AStA-Advances in Statistical Analysis 95(4), 325-529, 2011 erschienen.
Statistical Analysis of Longitudinal/Panel Data Structures Arising from Complex Interdisciplinary Questions
In all disciplines where statistical methods are applied to empirical data in order to model individual behaviour, there exists the problem of unobserved or unobservable individual heterogeneity, leading to biased results and hence unreliable inferences. This problem can be solved only by the availability of multiple observations on each individual that is, by the use of longitudinal, or panel data. In the light of both the rapid increase in data collection facilities and the rapid decrease in computational costs, the availability of panel data sets is now standard in many disciplines such as Sociology, Economics, Psychology, Medicine, Biology, etc. Interestingly, the underlying problem of inadequacy of statistical analyses and its remedy by using suitable methods on a richer data basis lead to many parallel and complimentary developments in the field sub-disciplines of mathematical statistics such as econometrics, biostatistics, psychometrics etc. and their respective interpretations. A by-product of this process can be seen in the development of novel interdisciplinary insights and approaches in several fields. Recent contributions exploit the structure of panel data to address issues such as the economical or psychological interpretation of social interactions of different peer groups formed with respect to sociological, demographical, ethnical, geographical, etc. structures. The empirically measurable manifestations of these interactions can be observed in the dynamic process of intra-individual actions and re-actions based on a cross-section of individual time series. At the same time longitudinal, or panel data are characterized by a number of remarkable advantages which make them superior to cross-sectional research: They allow the identification of causal relationships and provide information on intra-individual development, that is constancy and change within the individual across time. Through longitudinal analysis the constancy and change in relationship among other factors and covariates can be investigated. Repeated within-subjects designs permit the evaluation of the course and outcome of program interventions (e.g., programs which reintegrate unemployed workers into the workforce). Longitudinal data allow the study of critical life events or periods and their impact on the development of an individual. They allow the investigation of developmental trajectories or pathways in certain (sub-) populations. The workshop intends to demonstrate and promote the central role of panel data collection and analysis both in statistical and methodological developments. It aims at considering the different scientific disciplines where panel data are being used to explain substance matter questions. Of practical relevance is the fact that interdisciplinary longitudinal data require statistical methods which do not belong to the standard repertoire of curricula of graduate studies. Moreover, longitudinal or panel data encompass many methodological problems, which interfere with their numerous advantages. For instance, the missing data problem and consequently the imputation of estimated data are especially difficult in complex longitudinal designs. Longitudinal research may be done according to a wrong number and spacing of observations or they lead to higher attrition rates and drop-outs than cross-sectional studies. The aim of the workshop is to provide practical and up-to-date solutions to those methodological issues for the analysis of interdisciplinary longitudinal data and to present fascinating statistical applications.
Abstracts of Talks