Hidden Markov models: Estimation theory and economic applications
Leister, Anna Maria
In this thesis, maximum likelihood estimation of hidden Markov models in several settings is investigated. Nonparametric estimation of state-dependent general mixtures and log-concave densities is discussed theoretically and algorithmically. Penalized estimation for parametric hidden Markov models comparing several penalty functions is studied. In addition, various models based on mixture models and hidden Markov models differing in dependency structure and the inclusion of covariables are applied to a set of panel data containing the GDP of several countries.
Philipps-Universität Marburg
Mathematics
urn:nbn:de:hebis:04-z2016-01208
https://doi.org/10.17192/z2016.0120
opus:6661
https://archiv.ub.uni-marburg.de/diss/z2016/0120/cover.png
Mathematics
Mathematik
Publikationsserver der Universitätsbibliothek Marburg
Universitätsbibliothek Marburg
Hidden Markov models: Estimation theory and economic applications
EM-algorithm
urn:nbn:de:hebis:04-z2016-01208
Hidden Markov Modell
Fachbereich Mathematik und Informatik
Schätztheorie
Reine und Angewandte Mathematik
ths
Prof. Dr.
Holzmann
Hajo
Holzmann, Hajo (Prof. Dr.)
2016-05-02
https://doi.org/10.17192/z2016.0120
English
monograph
Philipps-Universität Marburg
Maximum Likelihood Schätzung
Leister, Anna Maria
Leister
Anna Maria
Hidden Markov Modelle: Schätztheorie und ökonomische Anwendungen
Die vorliegende Arbeit behandelt Maximum Likelihood Schätzung von Hidden Markov Modellen in unterschiedlichen Szenarien. Nichtparametrische Schätzung zustandsbedingter Mischungsmodelle und log-konkaver Dichten wird theoretisch und algorihmisch diskutiert. Penalisierte Schätzung für parametrische Hidden Markov Modelle unter unterschiedlichen Penalisierungsfunktionen wird untersucht. Diverse Modelle basierend auf Mischungsmodellen und Hidden Markov Modellen mit und ohne Kovariablen werden auf einen makroökonomischen Paneldatensatz zur Untersuchung von Einkommensverteilungen angewendet.
126
application/pdf
opus:6661
2016-05-25
doctoralThesis
In this thesis, maximum likelihood estimation of hidden Markov models in several settings is investigated. Nonparametric estimation of state-dependent general mixtures and log-concave densities is discussed theoretically and algorithmically. Penalized estimation for parametric hidden Markov models comparing several penalty functions is studied. In addition, various models based on mixture models and hidden Markov models differing in dependency structure and the inclusion of covariables are applied to a set of panel data containing the GDP of several countries.
EM-Algorithmus
2016
Statistik
estimation theory
statistics
Hidden Markov models
PRESERVATION_MASTER
VIEW
Image
PRESERVATION_MASTER