This dissertation studies various issues related to regime switching and DSGE models. The methods developed are used to study U.S. business cycles. Chapter one considers and derives the limit distributions of likelihood ratio based tests for Markov regime switching in multiple parameters in the context of a general class of nonlinear models. The analysis simultaneously addresses three difficulties: (1) some nuisance parameters are unidentified under the null hypothesis, (2) the null hypothesis yields a local optimum, and (3) the conditional regime probabilities follow stochastic processes that can only be represented recursively. When applied to US quarterly real GDP growth rates, the tests suggest strong evidence favoring the regime switching specification over a range of sample periods. Chapter two develops a modified likelihood ratio (MLR) test to detect regime switching in state space models. I apply the filtering algorithm introduced in Gordon and Smith (1988) to construct a modified likelihood function under the alternative hypothesis of two regimes and I extend the analysis in Chapter one to establish the asymptotic distribution of the MLR statistic under the null hypothesis of a single regime. I also apply the test to a simple model of the U.S. unemployment rate. This contribution is the first to develop a test based on the likelihood ratio principle to detect regime switching in state space models. The final chapter estimates a search and matching model of the aggregate labor market with sticky price and staggered wage negotiation. It starts with a partial equilibrium search and matching model and expands into a general equilibrium model with sticky price and staggered wage. I study the quantitative implications of the model. The results show that (1) the price stickiness and staggered wage structure are quantitatively important for the search and matching model of the aggregate labor market; (2) relatively high outside option payments to the workers, such as unemployment insurance payments, are needed to match the data; and (3) workers have lower bargaining power relative to firms, which contrasts with the assumption in the literature that workers and firms share equally the surplus generated from their employment relationship.

[pt] Este trabalho versa sobre previsão de reservas do tipo IBNR levando-se em conta uma ordenação diferente do triângulo de runoff incremental. Esta se dá por linhas empilhadas, originando, assim, uma série temporal univariada repleta de valores faltantes, cuja soma desses valores constitui o IBNR a ser estimado. Duas abordagens de estimação, inteiramente baseadas na teoria dos modelos em Espaço de Estado e do filtro de Kalman, são desenvolvidas, implementadas com dados reais de empresas seguradoras, e comparadas entre si e a outros métodos de estimação já consagrados na literatura atuarial. A primeira abordagem pauta-se no cálculo da matriz de covariâncias condicionais das componentes do IBNR, e a segunda é um processo de obtenção do IBNR por acumulação. Os resultados obtidos revelam, para as abordagens propostas, os seguintes pontos sumários: (i) plena eficiência e viabilidade computacional; (ii) sistemático ganho em termos de acurácia do IBNR estimado; e (iii) abrangência no que diz respeito às possibilidades de modelagem estatística dos dados de IBNR. / [en] This work deals with prediction of IBNR reserves under a different ordering of the non-cumulative runoff triangle. This is accomplished by stacking the rows, which results in a univariate time series with several missing values, whose corresponding sum is in fact the IBNR. Two estimation approaches, entirely based on state space methods and Kalman filtering, are developed, implemented with real data, and compared with some well established estimation methods for IBNR. The first approach consists in obtaining the conditional covariance matrix of the IBNR components, and the second tackles the IBNR estimation under an accumulation process. Three remarks emerge from the empirical results: (i)computational feasibility and efficiency; (ii)precision improvement for IBNR estimation; and (iii)flexibility in which concerns the IBNR modelling framework.

[pt] Os métodos de análise de séries temporais têm se revelado uma importante ferramenta de apoio à tomada de decisões, com importância crescente em um mundo cada vez mais globalizado. Esse fato pode ser ilustrado, entre muitos outros, através de um convênio firmado entre o CEPEL, o Núcleo de Estatística Computacional da PUC/RJ e a Eletrobrás, para se avaliar a utilidade dessas ferramentas nas etapas do planejamento do setor elétrico brasileiro. A metodologia em Espaço de Estado proporcionou o surgimento de duas importantes classes de modelos de previsão e análise de séries temporais completamente alternativas (os modelos estruturais e os modelos de inovações em espaço de estado), e, por isso, podem por vezes, causar dúvidas quando se fala em métodos de previsão em espaço de estado sem se especificar sobre qual das duas se está falando. Foi escolhido uma técnica específica e facilmente executável em softwares comerciais para cada classe de modelos: O desenvolvimento clássico de Harvey implementado no software STAMP, representando os modelos estruturais; e o desenvolvimento de Goodrich implementado no software FMP, representando os modelos de inovações. Essas técnicas estão tratadas de uma forma aprofundada, para proporcionar um melhor entendimento teórico das diferenças existentes entre ambas. Com o intuito de se avaliar a performance frente às outras técnicas existentes, são comparados os resultados das previsões entre as metodologias a partir de um sistema de comparação baseado nas estatísticas MAPE (Mean Absolute Percentage Error), RMSE (Root Mean Squared Error) e U-Theil. Para tanto são vistos sucintamente as técnicas: Alisamento Exponencial (Holt-Winters), Box & Jenkins e Redes Neurais. Todas as técnicas foram aplicadas aos dados de consumo de energia elétrica das 32 empresas concessionárias do setor no Brasil, além de comparadas com as previsões realizadas por essas concessionárias. A novidade deste trabalho para o projeto em andamento está na aplicação multivariada possível através da metodologia de Goodrich. / [en] The analysis of time series is, nowadays one of the most important tools in the decision making process, due mainly to the globalization of the world. As an illustration of that we can mention the recent contract signed between NEC/PUC-Rio and CEPEL/Eletrobrás, where time series techniques are to be used in the planning process of the brazilian sector. The state-space approach forms the basis of two important forecasting models to time series analysis the structural model and the state space innovation model. Because of that one finds it difficult to have a clear cut definition of either one of them. These two models formulation were implemented in comercial softwares: the structural model of A. Harvey in STAMP and the state space innovation of R. Goodrich in FMP. In order to check the perfomance of these state space approaches vis-à-vis the traditional forecasting techniques, it was used the following statistics: MAPE (Mean Absolute Percentage Error), RMSE (Root Mean Squared Error) and U-Theil. The traditional approaches used in the comparison were: Holt-Winters, Box & Jenkins and Backpropagation Neural Network. All the methods, included the state space ones were applied to the demand series of 32 electrical utilities which form the brazilian electrical distribution system. If was also attempted the multivariate state-space formulation of R. Goodrich which is included in FMP software.

[pt] Este trabalho consiste em se determinar técnicas de modelagem de bacias hidráulicas utilizando o conceito de espaço de estado, e escolha de um dos métodos que seja mais objetivo e prático para aplicação generalizando o estudo de Mendonza [1]. São estudados quatro métodos de obtenção do sistema em espaço de estado que são, método de expansão em frações parciais, programação interativa, modelo controlável e modelo observável. Pela característica de simplicidade de obtenção do modelo, foi escolhido este último que simplifica bastante a aplicação em bacias hidráulicas. Esta teoria é aplicada em seguida a duas bacias hidráulicas para obtenção dos modelos de sistema discretos em espaço de estado e são feitas simulações para as bacias do Rio Peixe e do Rio Grande. Os dados numéricos utilizados neste trabalho são hipotéticos. / [en] This thesis consists in determine moulding techiques of hydraulics basin using the state space representation, and choicing one of the most objective and practical methods for application, generalizing the study of Mendonza [1]. Are studied four methods for obtaining the system in state space which are, partial-fraction expansion method, interative programming, controllable model and observable model. By the simplicity characteristic to obtain the model, was chosen the last one to simplify enough the application in hydraulics basins. This theory is applied in sequence to two hydraulics basins for the obtaining of the models of discrete systems in state space and the simulations were done for the Rio do Peixe and Rio Grande basins. The numerical data used in this thesis are hypothetic.

[pt] No presente trabalho, é estudado por simulação o comportamento de um sistema de potência, constituído por duas áreas interligadas, diante da perda de geração em uma das áreas, utilizando-se técnicas de variáveis de estado, e considerando-se algumas não linearidades do mesmo. A partir do modelo representativo do sistema, [2], sob tratamento não linear, obtém-se um conjunto de equações, que representadas no espaço de estado, permitem a análise do comportamento do mesmo. O modelo que inclui o controlador do tipo integral, cujo objetivo é manter a freqüência gerada em 60 Hz, é de ordem doze, que após considerações de natureza simplificadora, se reduz à ordem seis. Determinam-se a seguir, os pontos de equilíbrio do sistema e, através do teorema de Liapunov, da estabilidade local, analisa-se a natureza dos mesmos. Como última etapa do trabalho, foi feita a simulação do modelo global do sistema, em computador digital, considerando-se diversas situações de perda de geração, caracterizando-se as áreas, por parâmetros diferentes, apresentando-se os resultados graficamente. / [en] In the present work, it is studied by simulation the dinamic behavior of a power system,constituted by two interconnected areas, in case of getting lost the power generation in one of two areas, by the use of technics in state variables and considering some non linearity of the system. Starting in the system representative model, [2], under non linear treatment, we can get a collection of equations that, if represented in the state space, will allow the analyse of its behavior. The model that include the controller of the integral type, which objective is to keep the generated frequency in 60 Hz, is of the order 12, that, after considerations of simplifyting nature, is reduced to the order six. The next step is to determine the balanced points of the system and by using the Liapunov`s theorem of the local stability, will be analysed the nature (Kind) of them. As the last step of the work, was made the simulation of the system global model using a digital computer and considering some types of generation lost situations, defining the areas by the use of different parameters and the results are shown graphically.

abstract: Transfer learning refers to statistical machine learning methods that integrate the knowledge of one domain (source domain) and the data of another domain (target domain) in an appropriate way, in order to develop a model for the target domain that is better than a model using the data of the target domain alone. Transfer learning emerged because classic machine learning, when used to model different domains, has to take on one of two mechanical approaches. That is, it will either assume the data distributions of the different domains to be the same and thereby developing one model that fits all, or develop one model for each domain independently. Transfer learning, on the other hand, aims to mitigate the limitations of the two approaches by accounting for both the similarity and specificity of related domains. The objective of my dissertation research is to develop new transfer learning methods and demonstrate the utility of the methods in real-world applications. Specifically, in my methodological development, I focus on two different transfer learning scenarios: spatial transfer learning across different domains and temporal transfer learning along time in the same domain. Furthermore, I apply the proposed spatial transfer learning approach to modeling of degenerate biological systems.Degeneracy is a well-known characteristic, widely-existing in many biological systems, and contributes to the heterogeneity, complexity, and robustness of biological systems. In particular, I study the application of one degenerate biological system which is to use transcription factor (TF) binding sites to predict gene expression across multiple cell lines. Also, I apply the proposed temporal transfer learning approach to change detection of dynamic network data. Change detection is a classic research area in Statistical Process Control (SPC), but change detection in network data has been limited studied. I integrate the temporal transfer learning method called the Network State Space Model (NSSM) and SPC and formulate the problem of change detection from dynamic networks into a covariance monitoring problem. I demonstrate the performance of the NSSM in change detection of dynamic social networks. / Dissertation/Thesis / Doctoral Dissertation Industrial Engineering 2015