Missing values in correlated observations (Q2768500)

Correlated observations are met with clinical studies of longitudinal data, panel data in econo\-mi\-co-scientific or socio-demographic researches and in the area of spatial statistics. Missing data due to non-attainability, non-response, unforseen mishaps, non-applicability, etc., are endemic problems in statistical samples and surveys. However this circumstance is usually ignored, so that statistical analyses based on reducing data lead to wrong or at least inefficient results. This booklet is devoted to the solution of these problems by means of Generalized Estimating Equations (GEE) which are the starting point for the Weighted Estimating Equations (WEE). Different strategies for regulation of the weights are discussed. An extension of the WEE to any dependence structure is proposed, and application to the treatment of missing values in various fields is presented.NEWLINENEWLINEContents: Preface. Marginal Regression models for Complete Data. Bases of the Treatment of Missing Values. Missing Values in Longitudinal Data. Weighted Estimating Equations. Weighted Estimating Equations for Dependence Structures. Choice Models and Diagnostic Models. Applied Examples. Synopses and Outlook. Estimation Algorithms for GEE and WEE. Calculations. Tables of results of simulation studies. Software. Literature.