This volume containes most of the scientific contributions presented at the First International Workshop on Correlated Data Modeling, held in Trieste, October 22-23, 1999. The University of Trieste has a notable tradition about the topic of correlation and data dependencies. Indeed, research on this topic dates back to Bruno de Finetti, who was professor at this university from 1947 to 1955. De Finetti introduced the notion of exchangeability in 1928 and that of partial exchangeability in 1937 and developed these concepts during his stay in Trieste. Both schemes have a great importance in modern statistics, representing a generalization of concepts of wide applicability, like Bayesian hierarchical models.The initiative has been promoted by the organizers in view of the growing interest that the treatment of correlation in modeling real data has gained in the scientific (not only statistical) literature. Although the main focus of the Workshop was on Generalized Estimating Equations as an approach for modeling correlated data, a wide perspective has been kept to allow contributions related to alternative approaches. The volume is organized in five sections. In the first section a short, selective survey on the estimating equations approach and further important generalizations are presented. The second section deals with the fundamental problem of treating missing data in longitudinal studies. New models and approaches are presented in the third section, where the issues of model selection and data simulation are discussed. The fourth section is dedicated to novel applications of Generalized Estimating Equations and related models in different fields (from political decision making to medicine and biology). Section five includes several contributions about alternative approaches that allow us to account for correlation and also pinpoints some open issues in the field. We hope that this volume will help to promote the application of estimating equations in various field of application, and motivates further comparisons with alternative methods. Moreover, we hope that this workshop will demonstrate (once again) the increasing need of exchanging ideas among different disciplines, like biometrics or econometrics, as most of the methods presented here experienced a separate (although intense) development.
The Editors
Contents General Remarks
M. Crowder , Estimating Equations: a Short, Selective Survey
G. MacKenzie, J. Reeves , Modelling Bivariate Longitudinal Data with Serial Correlation
M. Pourahrnadi , Joint Modeling of Mean, Dispersion and Correlation: Generalized Gaussian Estimating Equations
Missing Data
G. M. Fitzmaurice , Adjusting for Nonignorable Dropouts in Longitudinal Studies: A Mixture Modelling Approach Using GEE
M. C. Paik, R. Sacco, I-Feng Lin , Bivariate Binary Data Analysis with Non-Ignorably Missing Outcomes
Modeling
C. Lange, J. C. Whittaker , Model Selection for Quasi-Likelihood: A New Approach
T. Reilly , Modeling Correlated Data Using the Multivariate Normal Copula
Applications
C. J. W. Zorn , GEE Models of Political Decision Making
A. Ziegler, C. Kastner , The Effect of Misspecified Response Probabilities on Parameter Estimates from Weighted Estirnating Equations
M. A. Mugglestone, M. G. Kenward, S. J. Clark , Generalized Estimating Equations for Spatially Referenced Binary Data
M. Baccini, L. Lusa, M. Farella, A. Michelotti , Analysis of Repeated Measurements with Generalized Estimating Equations and Model Checking: an Electromyographic Study of the Human Jaw Muscles
Other Approaches and Open Issues
Castrignanò, L. Giglio, M. Stelluti , Application of Non-Parametric Geostatistics Agriculture
A. Di Bartolo , Human Capital Estimation through Structural Equation Models with some Categorical Observed Variables
A. Durio , A Model of Structural Equations for Quality Evaluation in the Medical Field
M. Cazzaro, R. Colombi , A Hybrid Parameterization for Contingency Tables
R. Ignaccolo , Kernel Estimation for Investment Funds' Evaluation
E. D. Isaia , Gauging Scores Control Charts: Some Remarks on a Two Pairs of Gauging Scheme
K. Moder, P. Parasiewicz , Statistical Comparison of Physical Habitat Sampling Strategies in Streams
This volume containes most of the scientific contributions presented at the First International Workshop on Correlated Data Modeling, held in Trieste, October 22-23, 1999. The University of Trieste has a notable tradition about the topic of correlation and data dependencies. Indeed, research on this topic dates back to Bruno de Finetti, who was professor at this university from 1947 to 1955. De Finetti introduced the notion of exchangeability in 1928 and that of partial exchangeability in 1937 and developed these concepts during his stay in Trieste. Both schemes have a great importance in modern statistics, representing a generalization of concepts of wide applicability, like Bayesian hierarchical models.The initiative has been promoted by the organizers in view of the growing interest that the treatment of correlation in modeling real data has gained in the scientific (not only statistical) literature. Although the main focus of the Workshop was on Generalized Estimating Equations as an approach for modeling correlated data, a wide perspective has been kept to allow contributions related to alternative approaches. The volume is organized in five sections. In the first section a short, selective survey on the estimating equations approach and further important generalizations are presented. The second section deals with the fundamental problem of treating missing data in longitudinal studies. New models and approaches are presented in the third section, where the issues of model selection and data simulation are discussed. The fourth section is dedicated to novel applications of Generalized Estimating Equations and related models in different fields (from political decision making to medicine and biology). Section five includes several contributions about alternative approaches that allow us to account for correlation and also pinpoints some open issues in the field. We hope that this volume will help to promote the application of estimating equations in various field of application, and motivates further comparisons with alternative methods. Moreover, we hope that this workshop will demonstrate (once again) the increasing need of exchanging ideas among different disciplines, like biometrics or econometrics, as most of the methods presented here experienced a separate (although intense) development.
