Bayesian analysis of binary panel probit models. The case of measurement error and missing values in explaining factors.

Autor/inn/en	Aßmann, Christian; Goßmann, Solange; Schönberger, Benno
Institution	Leibniz-Institut für Bildungsverläufe
Titel	Bayesian analysis of binary panel probit models. The case of measurement error and missing values in explaining factors.
Quelle	Bamberg: Leibniz Institute for Educational Trajectories (LIfBi) (2014), 27 S. PDF als Volltext kostenfreie Datei Link als defekt meldenVerfügbarkeit
Reihe	NEPS working paper. 35
Beigaben	Literaturangaben
Sprache	englisch
Dokumenttyp	online; Monographie; Graue Literatur
Schlagwörter	Schülerbefragung; Fehleranalyse; Datenanalyse; Regressionsanalyse; Markowsche Kette; Messgenauigkeit; Messfehler; NEPS (National Educational Panel Study)
Abstract	Since large panel data sets, e.g. on educational or epidemiological issues, are despite tremendous efforts in field work almost inevitably plagued by missing data and measurement error, the development of appropriate estimation techniques is necessary. Bayesian analysis facilitated via Markov Chain Monte Carlo (MCMC) sampling algorithms allows for conceptually straightforward treatment of measurement error and missing values based on the device of data augmentation. Augmenting the parameter vector by the missing values allows for direct incorporation of the uncertainty stemming from missing values into parameter estimation. Full conditional distributions for missing values are provided on the basis of a nonparametric sequential regression modeling approach. For empirical illustration the proposed methodology is applied for students participating in a survey of the National Educational Panel Study (NEPS) assessing the impact of curricular reforms. The empirical application points at the necessity to cope with missing data and measurement errors in order to avoid biased estimation. Additionally a simulation study is performed documenting the adequacy of the proposed estimation methodology. (Orig.).
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2020/3