A nonparametric multiple imputation approach for multilevel filtered questionnaires.

Autor/inn/en	Aßmann, Christian; Würbach, Ariane; Goßmann, Solange; Geissler, Ferdinand; Biedermann, Anika
Institution	Leibniz-Institut für Bildungsverläufe
Titel	A nonparametric multiple imputation approach for multilevel filtered questionnaires.
Quelle	Bamberg: Leibniz Institute for Educational Trajectories (LIfBi) (2014), 27 S. PDF als Volltext kostenfreie Datei Link als defekt meldenVerfügbarkeit
Reihe	NEPS working paper. 36
Beigaben	Literaturangaben
Sprache	englisch
Dokumenttyp	online; Monographie; Graue Literatur
Schlagwörter	Datenerhebung; Fragebogen; Filterverfahren; Datenanalyse; Regressionsanalyse; Variable; Messgenauigkeit; Einkommen; Erwachsener; NEPS (National Educational Panel Study); Deutschland
Abstract	Despite high efforts in field work and questionnaire design, low rates of missing values inevitably occur. The principles of multiple imputation allow for addressing this issue enhancing the analytical potential of the surveyed data. Large scale surveys provide rich data structures characterized by manifold discrete variables in combination with multilevel filtering in questionnaires. This requires multiple imputation techniques to preserve possible nonlinear relationships among the surveyed variables and full conditional distributions incorporating the information from multilevel filtering rules on an individual basis. To meet these requirements, a tree-based sequential regression approach is adapted addressing both the issues of possibly nonlinear relationships between categorical variables and complex multilevel filtering. Handling of filters within imputation is thereby adapted in a way to ensure consistency of the sequence of full conditional distributions. The suggested approach is illustrated in the context of income imputation in the adult cohort of the National Educational Panel Study. (Orig.).
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2020/3