Sessões Temáticas

ST1 - GAMLSS

Coordenador: Mário de Castro (ICMC/USP)

Palestrante: Mikis Stasinopoulos (Londonmet)

Título: GAMLSS Applications

Resumo: Since their introduction (Rigby and Stasinopoulos, 2005), GAMLSS models have been used in a variety of applied elds. For example GAMLSS have been applied among other elds: i) to the construction of reference charts for child growth curves Borghi et al., (2006), Cole et al. (2009), ii) to the analysis of ood frequencies (Villarini et al., 2009), iii) for normalising cDNA mi- croarray data (Khondoker et al., 2009), iv) on the health impact of temperatures in dwellings (Rudge and Gilchrist, 2007), v) to long-term rainfall data (Villarini et al., 2010), vi) to long-term survival models for clinical studies (de Castro et al., 2010), vii) to to investigate childhood obesity, Beyerlein et al. (2008). In this talk the speaker will describe his experience in applying GAMLSS to three
dierent problems:

• Oil production

• Long tail distributions

• time series data

This hopefully would give an indication of the exibility of GAMLS and how it could be adapted to deal with real data situations.

Palestrante: Raydonal Ospina (UFPE)

Título: An idea on parametric GAMLSS based on distances.

Resumo: In the context of parametric regression models, we propose a methodology that links a distance-based model and parametric GAMLSS models. An application is presented using the Gower distance.

ST2 - Previsão de Séries Temporais

Coordenador: Marcelo Cunha Medeiros (PUC/RJ)

Palestrante: Tore Schweder (U. Oslo)

Título: Causal sufficiency and Markov Completeness

Resumo: According to the probabilistic version of Hume's view of causality, real causal processes are Markovian. Granger causal relations between components of a Markov process are equivalent to local (in)dependencies. A sub-process of a Markov process is itself Markovian if and only if it is composed of components at the root of the directed local independence graph of the mother process. This is proved for a large class of Markov processes. Whether a model for time series data is causally sufficient, i.e. confounders or other important variables are not omitted, can thus be diagnostically tested by seeing whether the data accords with the proposed model being Markovian.

Palestrante: Flávio Ziegelmann (UFRGS)

Título: Multi-Period Volatility Predictions: A Comparative Study Using MIDAS Regressions

Resumo: We explore comparative results in the context of multi-period volatility predictions, focusing on the MIDAS approach. First, we compare the MIDAS method with two widely used methods of producing multi-period forecasts: the direct and the iterated approaches. Their relative performances are investigated in a Monte Carlo study and in a rst empirical study, where we predict volatility at horizons up to 60 days ahead. The results of the Monte Carlo study indicate that the MIDAS predictions are the most accurate at horizons of 15 days ahead and higher. The iterated forecasts are the best ones for shorter horizons (5 and 10 days ahead). In the rst empirical study, using daily returns of the S&P 500 and NASDAQ indexes, the results are not so conclusive, but suggest a better performance for the iterated forecasts. Second, we concentrate on relatively shorter horizons (up to 20 days ahead) to predict future volatility. In the second empirical study, we compare the MIDAS regressions with the HAR regressions and compute forecast combinations of the models. Using intra-daily data of the Bovespa index, we calculate a set of volatility measures to be used as regressors. In this case, the main result is that the combined
predictions signicantly outperform the individual ones in most cases. All analyses are out-of-sample.

Palestrante: Marcelo Cunha Medeiros (PUC/RJ)

Título: Estimating High-Dimensional Time Series Models

Resumo: We study the asymptotic properties of the adaptive LASSO (adaLASSO) in sparse, high-dimensional, linear time series models when both the number of covariates in the model and candidate variables can increase with the sample size, possibly being larger than the number of observations. We show, under mild conditions, that the adaLASSO correctly chooses the relevant variables as the number of observations increases and that it has the oracle property. We also derive an estimation algorithm based on the local quadratic approximation. Finally, we consider two empirical applications.

ST3 - Limites em Escala para Processos Estocásticos

Coordenador: Glauco Valle da Silva Coelho (UFRJ)

Palestrante: Milton Jara (IMPA)

Título: Additive functionals of one-dimensional particle systems.

Resumo: We introduce what we call second-order Boltzmann-Gibbs principle, and as an application we obtain the scaling limit of additive functionals of one-dimensional particle systems. Joint work with Patricia Gonçalves.

Palestrante: Fábio Julio Silva Valentim (UFES)

Título: Behavior of hydrodynamic models with conductances.

Resumo: For each function W, strictly increasing and right continuous, we assign a second order differential operator. The associated parabolic equation models, for instance, particles in an environment with permeable membranes. Such membranes, associated with the jump points of W, tend to reflect particles, creating spaces of discontinuity in the density profiles. This model can be formally established through hydrodynamic limit of interacting particle systems via an exclusion process with conductances induced by the function W. In recent works, models in one and higher dimensions were considered. In both cases, because of specific properties of the model, it was necessary to investigate analytical properties of such operators. In particular, stochastic homogenization results were obtained. These results allowed, beyond the proof of hydrodynamic limit, to obtain the equilibrium fluctuations for the model. We will present an overview of these facts and future challenges in this line.

Palestrante: Glauco Valle da Silva Coelho (UFRJ)

Título: Convergence to the Brownian Web for a generalization of the drainage
network model.

Resumo: We introduce a system of one-dimensional coalescing nonsimple random walks with long range jumps allowing crossing paths and exibiting dependence before coalescence. We show that under diffusive scaling this system converges in distribution to the Brownian Web. Joint work with Cristian Coletti.

ST4 - O Papel da Matemática (especialmente da Estatística) na Moderna Pesquisa sobre o Cancêr

Coordenador: Eduardo Jordão Neves (USP)

Palestrante: Luiz Fernando Lima Reis (Hospital Sírio Libanês)

Palestrante: Roger Chammas (USP)

Palestrante: Eduardo Jordão Neves (USP)