Apresentar-se-á noções teóricas de algumas das mais freqüentemente utilizadas análises de dados categorizados completos ou com omissão (i.e., que possuem dados incompletos ou faltantes) e ilustrar-se-á como realizar tais análises com a ajuda da biblioteca de rotinas Catdata implementada para o ambiente estatístico R.

• Accounting for Model Uncertainty via Trans-dimensional Genetic Algorithms (Ricardo Ehlers)

We develop for regression models trans-dimensional genetic algorithms for the exploration of large model spaces. Our algorithms can be used in two different ways. The first possibility is to search the best model according to some criteria such as AIC or BIC. The second possibility is to use our algorithms to explore the model space, search for the most probable models and estimate their posterior probabilities. This is accomplished by the use of genetic operators embedded in a reversible jump Markov chain Monte Carlo algorithm in the model space with several chains. As these chains run simultaneously and learn from each other via the genetic operators, our algorithm efficiently explores the large model space and easily escapes local maxima regions common in the presence of highly correlated regressors. We illustrate the power of our trans-dimensional genetic algorithms with applications to two real data sets.

• A novel approach for Evolutionary Computation through clustering techniques (Leonardo Ramos Emmendorfer)

Evolutionary Computation, which is a branch of Computational Intelligence, is applicable to some classes of hard combinatorial optimization problems. It comprises bio-inspired algorithms, which are characterized by maintaining and recombinating a "population" of candidate solutions (also called "individuals"). The preservation of the diversity in the population has already been established as an important concern for Evolutionary Computation, and clustering techniques were, among others, successfully applied with this purpose. Another important aspect of the research on that field is the "linkage learning- the detection of the structure of the problem, avoiding the disruption of building blocks when new individuals are generated. One of the most effective approaches for that purpose is the adoption of statistical models for the population, which could represent the structure of the problem being solved. The most effective algorithm of this class is the Bayesian Optimization Algorithm (BOA), which adopts a Bayesian Network as the statistical model for the promising solutions. However, it would be desirable to avoid the adoption of such kind of model, since the learning process must be executed several times at each run of the optimization algorithm, and a high computational effort is required for learning the structure of the Bayesian Network. A novel approach is proposed, where clustering, for the first time, plays both roles of diversity preservation and linkage learning. Despite of its very simple architecture, the new clustering-based algorithm proposed achieves competitive results when applied to some benchmark problems, at a lower computational cost and with better scalability, when compared to BOA.

• Tutorial do Sweave (Paulo Ribeiro)

Um tutorial indicando passo a passo como criar seu primeiro arquivo Sweave e compilá-lo no R e LaTeX.

• Modelos de Covariância Espaço-Temporais Gaussianos (Alexandre Sousa da Silva)

Campos aleatórios espaço-temporais gaussianos são completamente especificados pelo vetor de médias e a matriz de covariância. O vetor de média é facilmente especificado já a matriz de covariância precisa ser positiva definida para que seja considerada válida. Nesta apresentação serão discutidas algumas possibilidades de funções de covariância válidas, aplicadas em dois conjuntos de dados. O primeiro é se referente à estoque de peixe da costa de Portugual e o segundo se refere à capacidade de armazenagem de água em um solo com citros.