This thesis tackles two testing problems from the fields of treatment evaluation and social network analysis. In both cases, the presence of a flexible unobserved component invalidates popular testing procedures that have been developed for versions of the respective model that do not allow for unobserved heterogeneity or that make very restrictive assumptions about the nature of the unobserved heterogeneity. I suggest valid testing procedures and illustrate their empirical bite by applying them to real data. Moreover, I analyze a general framework for testing for index sufficiency when the hypothesized index is unknown but estimable.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Das Dokument wird vom Publikationsserver der Universitätsbibliothek Mannheim bereitgestellt.