3 resultados para Kernel estimator and ROC-GLM methodology
em ArchiMeD - Elektronische Publikationen der Universität Mainz - Alemanha
Resumo:
In dieser interdisziplinären, translationswissenschaftlichen Studie wird die Integration von Curriculum und Evaluierung in der Dolmetscherausbildung theoretisch fundiert und im Rahmen einer Fallstudie empirisch untersucht. Dolmetschkompetenz wird als ein durch zweckgerechte und messgenaue (valid and reliable) Bewertungsmethoden dokumentiertes Ergebnis der Curriculumanwendung betrachtet. Definitionen, Grundlagen, Ansätze, Ausbildungs- und Lernziele werden anhand der Curriculumtheorie und Dolmetschwissenschaft beschrieben. Traditionelle und alternative Evaluierungsmethoden werden hinsichtlich ihrer Anwendbarkeit in der Dolmetscherausbildung erprobt. In der Fallstudie werden die Prüfungsergebnisse zweier Master-Studiengänge-MA Konferenzdolmetschen und MA Dolmetschen und Übersetzen-quantitativ analysiert. Die zur Dokumentation der Prüfungsergebnisse eingesetzte Bewertungsmethodik wird qualitativ untersucht und zur quantitativen Analyse in Bezug gesetzt. Die Fallstudie besteht aus 1) einer chi-square-Analyse der Abschlussprüfungsnoten getrennt nach Sprachkombination und Prüfungskategorie (n=260), 2) einer Umfrage unter den Jurymitgliedern hinsichtlich der Evaluierungsansätze, -verfahren, und -kriterien (n = 45; 62.22% Rücklaufrate); und 3) einer Analyse des ausgangssprachlichen Prüfungsmaterials ebenfalls nach Sprachkombination und Prüfungskategorie. Es wird nachgewiesen, dass Studierende im MA Dolmetschen und Übersetzen tendenziell schlechtere Prüfungsleistungen erbringen als Studierende im MA Konferenzdolmetschen. Die Analyseergebnisse werden jedoch als aussageschwach betrachtet aufgrund mangelnder Evaluierungsvalidität. Schritte zur Curriculum- und Evaluierungsoptimierung sowie ein effizienteres Curriculummodell werden aus den theoretischen Ansätzen abgeleitet. Auf die Rolle der Ethik in der Evaluierungsmethodik wird hingewiesen.
Resumo:
In this treatise we consider finite systems of branching particles where the particles move independently of each other according to d-dimensional diffusions. Particles are killed at a position dependent rate, leaving at their death position a random number of descendants according to a position dependent reproduction law. In addition particles immigrate at constant rate (one immigrant per immigration time). A process with above properties is called a branching diffusion withimmigration (BDI). In the first part we present the model in detail and discuss the properties of the BDI under our basic assumptions. In the second part we consider the problem of reconstruction of the trajectory of a BDI from discrete observations. We observe positions of the particles at discrete times; in particular we assume that we have no information about the pedigree of the particles. A natural question arises if we want to apply statistical procedures on the discrete observations: How can we find couples of particle positions which belong to the same particle? We give an easy to implement 'reconstruction scheme' which allows us to redraw or 'reconstruct' parts of the trajectory of the BDI with high accuracy. Moreover asymptotically the whole path can be reconstructed. Further we present simulations which show that our partial reconstruction rule is tractable in practice. In the third part we study how the partial reconstruction rule fits into statistical applications. As an extensive example we present a nonparametric estimator for the diffusion coefficient of a BDI where the particles move according to one-dimensional diffusions. This estimator is based on the Nadaraya-Watson estimator for the diffusion coefficient of one-dimensional diffusions and it uses the partial reconstruction rule developed in the second part above. We are able to prove a rate of convergence of this estimator and finally we present simulations which show that the estimator works well even if we leave our set of assumptions.
Resumo:
This thesis focuses on the controlled assembly of monodisperse polymer colloids into ordered two-dimensional arrangements. These assemblies, commonly referred to as colloidal monolayers, are subsequently used as masks for the generation of arrays of complex metal nanostructures on solid substrates.rnThe motivation of the research presented here is twofold. First, monolayer crystallization methods were developed to simplify the assembly of colloids and to produce more complex arrangements of colloids in a precise way. Second, various approaches to colloidal lithography are designed with the aim to include novel features or functions to arrays of metal nanostructures.rnThe air/water interface was exploited for the crystallization of colloidal monolayer architectures as it combines a two-dimensional confinement with a high lateral mobility of the colloids that is beneficial for the creation of high long range order. A direct assembly of colloids is presented that provides a cheap, fast and conceptually simple methodology for the preparation of ordered colloidal monolayers. The produced two-dimensional crystals can be transformed into non-close-packed architectures by a plasma-induced size reduction step, thus providing valuable masks for more sophisticated lithographic processes. Finally, the controlled co-assembly of binary colloidal crystals with defined stoichiometries on a Langmuir trough is introduced and characterized with respect to accessible configurations and size ratios. rnSeveral approaches to lithography are presented that aim at introducing different features to colloidal lithography. First, using metal-complex containing latex particles, the synthesis of which is described as well, symmetric arrays of metal nanoparticles can be created by controlled combustion of the organic material of the colloids. The process does not feature an inherent limit in nanoparticle size and is able to produce complex materials as will be demonstrated for FePt alloy particles. Precise control over both size and spacing of the particle array is presented. rnSecond, two lithographic processes are introduced to create sophisticated nanoparticle dimer units consisting of two crescent shaped nanostructures in close proximity; essentially by using a single colloid as mask to generate two structures simultaneously. Strong coupling processes of the parental plasmon resonances of the two objects are observed that are accompanied by high near-field enhancements. A plasmon hybridization model is elaborated to explain all polarization dependent shifts of the resonance positions. Last, a technique to produce laterally patterned, ultra-flat substrates without surface topographies by embedding gold nanoparticles in a silicon dioxide matrix is applied to construct robust and re-usable sensing architectures and to introduce an approach for the nanoscale patterning of solid supported lipid bilayer membranes. rn