Econometric analysis of heterogeneous treatment and network models

Sarnetzki, Florian

Preview

PDF
Sarnetzki.pdf - Published
Download (916kB)

URL:	https://madoc.bib.uni-mannheim.de/38817
URN:	urn:nbn:de:bsz:180-madoc-388174
Document Type:	Doctoral dissertation
Year of publication:	2015
Place of publication:	Mannheim
University:	Universität Mannheim
Evaluator:	Mammen, Enno
Date of oral examination:	12 May 2015
Publication language:	English
Institution:	Außerfakultäre Einrichtungen > Graduate School of Economic and Social Sciences - CDSE (Economics) School of Law and Economics > Statistik (Mammen)
Subject:	310 Statistics
Subject headings (SWD):	Nichtparametrischer Test , Asymptotik , Analyse latenter Strukturen , Maximum-Likelihood-Schätzung , Konsistenz <Stochastik> , Zentralität
Keywords (English):	treatment effects , unobserved heterogeneity , overidentification test , instrumental variables , generated regressors , wild bootstrap , teenage pregnancies , network analysis , stochastic block model , latent position model , centrality
Abstract:	My dissertation investigates commonly used testing and estimation procedures and extends these by taking into account more heterogeneity. In chapter 2, me and my co-author Andreas Dzemski provide a new overidentification test that allows for essential heterogeneity. In chapter 3, I prove weak consistency up to a measure preserving transformation for maximum-likelihood estimation of unobserved latent positions in a Euclidean space just based on observable information of the agent's linking behavior. In chapter 4, I propose a new measure of centrality which exploits the latent space structure and identifies agents who connect clusters.