terça-feira, 28 de abril de 2015

Presentations - Tuesday (26/05/15)

Industrial Statistics in Future Manufacturing
Authors: Murat Kulahci (TUD)
Speaker: Murat Kulahci
Abstract: The future manufacturing environment is becoming more complex and will generate vast amounts of high dimensional data with complex dependencies. The information embedded in this data could and should be used for process improvement studies. However multivariate statistical methods traditionally used in process improvement applications are usually inadequate to handle this type of data. New sensor technologies such as sound and image can be used for proactive interventions to ensure product quality and process stability. We expect a great deal of research opportunities in industrial statistics applications to take advantage of the abundance of sensory data. Some of these opportunities will be discussed during the presentation.

Livestock classification as resistant or susceptible to endoparasite infections
Authors: Marcus Alexandre Nunes (UFRN); Adriana Santana do Carmo; Cristiano Amancio Vieira Borges; Michel Marques Farah; Márcia Cristina de Azevedo Prata; Marco Antônio Machado;John Furlong; Wagner Arbe; Marcos Vinícius Gualberto Barbosa da Silva; Embrapa Gado de Leite
Speaker: Marcus Alexandre Nunes
Abstract: Infection by endoparasites have a huge impact on animal health. Cattle, for example, can lose weight or yield less milk if infected. However, while some animals are prone to be infected, other animals are naturally resistant to endoparasites infection. In this work, we are interested in classify dairy cattle in different groups according to their infection susceptibility. First, we assumed three different profiles for the infestation behaviour: naturally resistant, naturally susceptible, and acquired immunity. In order to identify which animal belongs to each group, we analyzed data from 376 F2 1/2 Gyr and 1/2 Holstein animals, whose infections were recorded during 28 weeks. We used a clustering method called KmL, capable of identifying different clusters in longitudinal data, in order to identify how many clusters are needed to partition the original data into meaningful groups. Our findings show that only the resistant and susceptible groups need to be considered. We also compare the results of different generalized linear models fitted to the data.

CUSUM Control charts to monitor series of Negative Binomial count data
Authors: Airlane Pereira Alencar (IME-USP), Linda Lee Ho (EPUSP), Orlando Yesid Esparza Albarracin (IME-USP)
Speaker: Airlane Alencar  
Abstract: To detect outbreaks of diseases in public health, several control charts have been proposed in the literature. In this context, the usual generalized linear model may be fitted for counts under a Negative Binomial distribution with a logarithm link function and the population size included as offset to model hospitalization rates. Different statistics are used to build CUSUM control charts to monitor daily hospitalizations and their performances are compared in simulation studies. The main contribution of the current paper is to consider different statistics based on transformations and deviance residual to build control charts to monitor counts with seasonality effects and evaluate all the assumptions of the monitored statistics. The monitoring of daily number of hospital admissions due to respiratory diseases for people aged over 65years old in the city São Paulo-Brazil is considered as an illustration of the current proposal.

Double generalized linear model from a Bayesian perspective applied to designed experiments
Authors: Afranio Vieira (UFSCar)
Speaker: Afranio Vieira
Abstract: Joint generalized linear models and double generalized linear models (DGLMs) were designed to model outcomes for which the variability can be explained using factors and/or covariates. When such factors operate, the usual normal regression models, which inherently exhibit constant variance, will under-represent variation in the data and hence may lead to erroneous inferences. For count and proportion data, such noise factors can generate a so-called overdispersion effect, and the use of binomial and Poisson models underestimates the variability and, consequently, incorrectly indicate significant effects. In this work, we propose a DGLM from a Bayesian perspective, focusing on the case of proportion data, where the overdispersion can be modeled using a random effect that depends on some noise factors. The posterior joint density function was sampled using Monte Carlo Markov Chain algorithms, allowing inferences over the model parameters. An application to a data set under experimental design is presented, and a comparison with frequentist for which it is shown that the Bayesian approach is quite feasible, even when limited prior information is available.

Nenhum comentário:

Postar um comentário