TY  - THES
AB  - Using a model selection approach, this thesis proposes a constructive data-and-theory-combined procedure to identify model structures in the framework of a linear simultaneous equations system based on observed data. A model structure is characterized by restrictions on the structural parameters. To identify these restrictions two issues have to be taken into account: the first is the problem of observational equivalence, i.e. different models may have an identical density function, henceforth data cannot differentiate such observationally equivalent models; the second is the identification of the restrictions on structural parameters. For the first problem we classify models into different observationally equivalent classes and give necessary and sufficient conditions for the uniqueness of observationally equivalent models. For the second problem we take an approach based on the information criterion and give a (strong) consistent criterion to identify the restrictions on the structural parameters. We apply this model selection criterion to cointegration systems and provide a unified approach to analyzing linear simultaneous equations systems and cointegration systems. The model selection criterion is also used to identify the encompassing relations among different structural models under mis-specification. Through constructive use of the model selection criterion, we may get the most parsimonious structural model that is compatible to the data among the models under investigation. However, all conclusions conducted from the model selection criterion are valid only asymptotically. Nevertheless, the relevance for practical applications of this criterion is demonstrated by some simulation studies.
DA  - 2002
KW  - Model selection
KW  - Ökonometrie
KW  - Structural models
KW  - Simultaneous equation models
KW  - Strukturelle Modelle
KW  - Modellselektion
LA  - eng
PY  - 2002
TI  - Econometric structural models : a model selection approach
UR  - https://nbn-resolving.org/urn:nbn:de:hbz:361-11030
Y2  - 2024-11-27T03:23:02
ER  -