Analyse von Längsschnittdaten mit fehlenden Werten: Grundlagen, Verfahren und Anwendungen.
|Other Titles:||Analysis of longitudinal data with missing values. Foundations, methods and applications.||Authors:||Spiess, Martin||1. Expert:||Engel, Uwe||2. Expert:||Huinink, Johannes||Abstract:||
The first part gives an overview over foundations of empirical social research and an introduction into the estimation of linear fixed and random effects panel models. In addition, the semi parametric estimation of binary panel models based on generalized estimating equations (GEE) is addressed. The standard GEE approach, where the covariance structure parameters are treated as nuisance parameters is then generalized to include estimating equations for both, mean and covariance structure parameters. This approach allows the estimation of simultaneous equations panel models with mixed continuous and ordered categorical outcomes which is discussed in detail. As a measure of the explanatory power of the model a pseudo-R^2 measure is developed and evaluated. In the second part, fundamental concepts important with respect to the analysis of data sets with missing values are introduced and discussed and various approaches and methods to compensate for missing data are reviewed. The method of multiple imputation and its application is treated in detail. The approaches and techniques proposed and discussed in the first two parts are tested and illustrated with the help of various simulation studies and examples, respectively.The last chapter deals with possibly time changing effects of variables that can be interpreted as social investments on variables that can be interpreted as subjective and objective gratification variables. The resulting two-equation panel model with mixed continuous and ordered categorical outcomes is estimated with the approach described in the first part based on a data set with missing values. To compensate for missing data, a mixed weighting and multiple imputation approach is adopted.
|Keywords:||Panel models, generalized estimating equations, pseudo-R square, missing data, weighting, multiple imputation, status inconsistency||Issue Date:||4-Feb-2004||URN:||urn:nbn:de:gbv:46-diss000012631||Institution:||Universität Bremen||Faculty:||FB8 Sozialwissenschaften|
|Appears in Collections:||Dissertationen|
checked on Sep 24, 2020
checked on Sep 24, 2020
Items in Media are protected by copyright, with all rights reserved, unless otherwise indicated.