No description available.

The hydrothermal scheduling problem aims to determine an operation strategy that produces generation targets for each power plant at each stage of the planning horizon. This strategy aims to minimize the expected value of the operation cost over the planning horizon, composed of fuel costs to operate thermal plants plus penalties for failure in load supply. The system state at each stage is highly dependent on the water inflow at each hydropower generator reservoir. This work focuses on developing a probabilistic model for the inflows that is suitable for a multistage stochastic algorithm that solves the hydrothermal scheduling problem. The probabilistic model that governs the inflows is based on a dynamic linear model. Due to the cyclical behavior of the inflows, the model incorporates seasonal and regression components. We also incorporate climate variables such as precipitation, El Ni\~no, and other ocean indexes, as predictive variables when relevant. The model is tested for the power generation system in Brazil with about 140 hydro plants, which are responsible for more than 80\% of the electricity generation in the country. At first, these plants are gathered by basin and classified into 15 groups. Each group has a different probabilistic model that describes its seasonality and specific characteristics. The inflow forecast derived with the probabilistic model at each stage of the planning horizon is a continuous distribution, instead of a single point forecast. We describe an algorithm to form a finite scenario tree by sampling from the inflow forecasting distribution with interstage dependency, that is, the inflow realization at a specific stage depends on the inflow realization of previous stages. / text

We define a subclass of dynamic linear models with unknown hyperparameters called d-inverse-gamma models. We then approximate the marginal p.d.f.s of the hyperparameter and the state vector by the data augmentation algorithm of Tanner/Wong. We prove that the regularity conditions for convergence hold. A sampling based scheme for practical implementation is discussed. Finally, we illustrate how to obtain an iterative importance sampling estimate of the model likelihood. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

In the present paper we discuss Bayesian estimation of a very general model class where the distribution of the observations is assumed to depend on a latent mixture or switching variable taking values in a discrete state space. This model class covers e.g. finite mixture modelling, Markov switching autoregressive modelling and dynamic linear models with switching. Joint Bayesian estimation of all latent variables, model parameters and parameters determining the probability law of the switching variable is carried out by a new Markov Chain Monte Carlo method called permutation sampling. Estimation of switching and mixture models is known to be faced with identifiability problems as switching and mixture are identifiable only up to permutations of the indices of the states. For a Bayesian analysis the posterior has to be constrained in such a way that identifiablity constraints are fulfilled. The permutation sampler is designed to sample efficiently from the constrained posterior, by first sampling from the unconstrained posterior - which often can be done in a convenient multimove manner - and then by applying a suitable permutation, if the identifiability constraint is violated. We present simple conditions on the prior which ensure that this method is a valid Markov Chain Monte Carlo method (that is invariance, irreducibility and aperiodicity hold). Three case studies are presented, including finite mixture modelling of fetal lamb data, Markov switching Autoregressive modelling of the U.S. quarterly real GDP data, and modelling the U .S./U.K. real exchange rate by a dynamic linear model with Markov switching heteroscedasticity. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

O setor agrícola é, historicamente, um dos pilares da economia brasileira, e apesar de ter sua importância diminuída com o desenvolvimento do setor industrial e de serviços ainda é responsável por dar dinamismo econômico ao país, bem como garantir a segurança alimentar, auxiliar no controle da inflação e na formação de reservas monetárias. Neste contexto as safras agrícolas exercem grande influência no comportamento do setor e equilíbrio no mercado agrícola. Foram desenvolvidas diversas metodologias de previsão de safra, sendo em sua maioria modelos de simulação de crescimento. Entretanto, recentemente os modelos estatísticos vem sendo utilizados mais comumente devido às suas predições mais rápidas em períodos anteriores à colheita. No presente trabalho foram avaliadas duas destas metodologias, os modelos ARIMA e os Modelos Lineares Dinâmicos (MLD), sendo utilizada tanto a inferência clássica quanto a bayesiana. A avaliação das metodologias deu-se por meio da análise das previsões dos modelos, bem como da facilidade de implementação e poder computacional necessário. As metodologias foram aplicadas a dados de produção de soja para o município de Mamborê-PR, no período de 1980 a 2013, sendo área plantada (ha) e precipitação acumulada (mm) variáveis auxiliares nos modelos de regressão dinâmica. Observou-se que o modelo ARIMA (2,1,0) reparametrizado na forma de um MLD e estimado por meio de máxima verossimilhança, gerou melhores previsões do que aquelas obtidas com o modelo ARIMA(2,1,0) não reparametrizado. / The agriculture is, historically, one of Brazil\'s economic pillars, and despite having it\'s importance diminished with the development of the industry and services it still is responsible for giving dynamism to the country inland\'s economy, ensuring food security, controlling inflation and assisting in the formation of monetary reserves. In this context the agricultural crops exercise great influence in the behaviour of the sector and agricultural market balance. Diverse crop forecast methods were developed, most of them being growth simulation models, however, recently the statistical models are being used due to its capability of forecasting early when compared to the other models. In the present thesis two of these methologies were evaluated, ARIMA and Dynamic Linear Models, utilizing both classical and bayesian inference. The forecast accuracy, difficulties in the implementation and computational power were some of the caracteristics utilized to assess model efficiency. The methodologies were applied to Soy production data of Mamborê-PR, in the 1980-2013 period, also noting that planted area (ha) and cumulative precipitation (mm) were auxiliary variables in the dynamic regression. The ARIMA(2,1,0) reparametrized in the DLM form and adjusted through maximum likelihood generated the best forecasts, folowed by the ARIMA(2,1,0) without reparametrization.

Dynamic Linear Models are a state space model framework based on the Kalman filter. We use this framework to do seasonal adjustments of empirical and artificial data. A simple model and an extended model based on Gibbs sampling are used and the results are compared with the results of a standard seasonal adjustment method. The state space approach is then extended to discuss direct and indirect seasonal adjustments. This is achieved by applying a seasonal level model with no trend and some specific input variances that render different signal-to-noise ratios. This is illustrated for a system consisting of two artificial time series. Relative efficiencies between direct, indirect and multivariate, i.e. optimal, variances are then analyzed. In practice, standard seasonal adjustment packages do not support optimal/multivariate seasonal adjustments, so a univariate approach to simultaneous estimation is presented by specifying a Holt-Winters exponential smoothing method. This is applied to two sets of time series systems by defining a total loss function that is specified with a trade-off weight between the individual series’ loss functions and their aggregate loss function. The loss function is based on either the more conventional squared errors loss or on a robust Huber loss. The exponential decay parameters are then estimated by minimizing the total loss function for different trade-off weights. It is then concluded what approach, direct or indirect seasonal adjustment, is to be preferred for the two time series systems. The dynamic linear modeling approach is also applied to Swedish political opinion polls to assert the true underlying political opinion when there are several polls, with potential design effects and bias, observed at non-equidistant time points. A Wiener process model is used to model the change in the proportion of voters supporting either a specific party or a party block. Similar to stock market models, all available (political) information is assumed to be capitalized in the poll results and is incorporated in the model by assimilating opinion poll results with the model through Bayesian updating of the posterior distribution. Based on the results, we are able to assess the true underlying voter proportion and additionally predict the elections. / <p>At the time of doctoral defence the following papers were unpublished and had a status as follows: Paper 3: Manuscript; Paper 4: Manuscripts</p>

O setor agrícola é, historicamente, um dos pilares da economia brasileira, e apesar de ter sua importância diminuída com o desenvolvimento do setor industrial e de serviços ainda é responsável por dar dinamismo econômico ao país, bem como garantir a segurança alimentar, auxiliar no controle da inflação e na formação de reservas monetárias. Neste contexto as safras agrícolas exercem grande influência no comportamento do setor e equilíbrio no mercado agrícola. Foram desenvolvidas diversas metodologias de previsão de safra, sendo em sua maioria modelos de simulação de crescimento. Entretanto, recentemente os modelos estatísticos vem sendo utilizados mais comumente devido às suas predições mais rápidas em períodos anteriores à colheita. No presente trabalho foram avaliadas duas destas metodologias, os modelos ARIMA e os Modelos Lineares Dinâmicos (MLD), sendo utilizada tanto a inferência clássica quanto a bayesiana. A avaliação das metodologias deu-se por meio da análise das previsões dos modelos, bem como da facilidade de implementação e poder computacional necessário. As metodologias foram aplicadas a dados de produção de soja para o município de Mamborê-PR, no período de 1980 a 2013, sendo área plantada (ha) e precipitação acumulada (mm) variáveis auxiliares nos modelos de regressão dinâmica. Observou-se que o modelo ARIMA (2,1,0) reparametrizado na forma de um MLD e estimado por meio de máxima verossimilhança, gerou melhores previsões do que aquelas obtidas com o modelo ARIMA(2,1,0) não reparametrizado. / The agriculture is, historically, one of Brazil\'s economic pillars, and despite having it\'s importance diminished with the development of the industry and services it still is responsible for giving dynamism to the country inland\'s economy, ensuring food security, controlling inflation and assisting in the formation of monetary reserves. In this context the agricultural crops exercise great influence in the behaviour of the sector and agricultural market balance. Diverse crop forecast methods were developed, most of them being growth simulation models, however, recently the statistical models are being used due to its capability of forecasting early when compared to the other models. In the present thesis two of these methologies were evaluated, ARIMA and Dynamic Linear Models, utilizing both classical and bayesian inference. The forecast accuracy, difficulties in the implementation and computational power were some of the caracteristics utilized to assess model efficiency. The methodologies were applied to Soy production data of Mamborê-PR, in the 1980-2013 period, also noting that planted area (ha) and cumulative precipitation (mm) were auxiliary variables in the dynamic regression. The ARIMA(2,1,0) reparametrized in the DLM form and adjusted through maximum likelihood generated the best forecasts, folowed by the ARIMA(2,1,0) without reparametrization.

Pricing decision support systems have been developed in order to help retail companies optimise the prices they set when selling their goods and services. This research aims to enhance the essential forecasting and optimisation techniques that underlie these systems. This is first done by applying the method of Dynamic Linear Models in order to provide sales forecasts of a higher accuracy compared with current methods. Secondly, the method of Support Vector Regression is used to forecast future competitor prices. This new technique aims to produce forecasts of greater accuracy compared with the assumption currentlyused in pricing decision support systems that each competitor's price will simply remain unchanged. Thirdly, when competitor prices aren't forecasted, a new pricing optimisation technique is presented which provides the highest guaranteed profit. Existing pricing decision support systems optimise price assuming that competitor prices will remain unchanged but this optimisation can't be trusted since competitor prices are never actually forecasted. Finally, when competitor prices are forecasted, an exhaustive search of a game-tree is presented as a new way to optimise a retailer's price. This optimisation incorporates future competitor price moves, something which is vital when analysing the success of a pricing strategy but is absent from current pricing decision support systems. Each approach is applied to the forecasting and optimisation of daily retail vehicle fuel pricing using real commercial data, showing the improved results in each case.

Clinical drug trials are costly and time-consuming. Bayesian methods alleviate the inefficiencies in the testing process while providing user-friendly probabilistic inference and predictions from the sampled posterior distributions, saving resources, time, and money. We propose a dynamic linear model to estimate the mean response at each dose level, borrowing strength across dose levels. Our model permits nonmonotonicity of the dose-response relationship, facilitating precise modeling of a wider array of dose-response relationships (including the possibility of toxicity). In addition, we incorporate an adaptive approach to the design of the clinical trial, which allows for interim decisions and assignment to doses based on dose-response uncertainty and dose efficacy. The interim decisions we consider are stopping early for success and stopping early for futility, allowing for patient and time savings in the drug development process. These methods complement current clinical trial design research.

High dimensional space-time datasets are available nowadays in various aspects of life such as economy, agriculture, health, environment, etc. Meanwhile, it is challenging to reveal possible connections between climate change and weather extreme events such as hurricanes or tornadoes. In particular, the relationship between tornado occurrence and climate change has remained elusive. Moreover, modeling multivariate spatio-temporal data is computationally expensive. There is great need to computationally feasible models that account for temporal, spatial, and inter-variables dependence. Our research focuses on those areas in two ways. First, we investigate connections between changes in tornado risk and the increase in atmospheric instability over Oklahoma. Second, we propose two multiscale spatio-temporal models, one for multivariate Gaussian data, and the other for matrix-variate Gaussian data. Those frameworks are novel additions to the existing literature on Bayesian multiscale models. In addition, we have proposed parallelizable MCMC algorithms to sample from the posterior distributions of the model parameters with enhanced computations. / Doctor of Philosophy / Over 1000 tornadoes are reported every year in the United States causing massive losses in lives and possessions according to the National Oceanic and Atmospheric Administration. Therefore, it is worthy to investigate possible connections between climate change and tornado occurrence. However, there are massive environmental datasets in three or four dimensions (2 or 3 dimensional space, and time), and the relationship between tornado occurrence and climate change has remained elusive. Moreover, it is computationally expensive to analyze those high dimensional space-time datasets. In part of our research, we have found a significant relationship between occurrence of strong tornadoes over Oklahoma and meteorological variables. Some of those meteorological variables have been affected by ozone depletion and emissions of greenhouse gases. Additionally, we propose two Bayesian frameworks to analyze multivariate space-time datasets with fast and feasible computations. Finally, our analyses indicate different patterns of temperatures at atmospheric altitudes with distinctive rates over the United States.