A SBio-SPE e a SGAPEIO (Sociedade Galega para a Promoción da Estatistica e da Investigación de Operacións) pretendem iniciar o ano de 2025 com o I Simpósio Online Galaico-Portuguesa. O evento irá decorrer no próximo dia 17/01/2025, entre as 14:30h–16:30h WET (+1h CET) e contará com a participação de 4 oradores:
Informações e Registo
O Simpósio irá decorrer a 17/01/2025, sexta-feira, entre as 14:30h e as 16:30h (hora de Portugal; +1 hora de Galiza).
A participação é GRATUITA para todos, inclusive não-sócios da SPE/SGAPEIO. Contudo o registo é obrigatório para que a sessão possa ser disponibilizada.
O registo no Simpósio pode ser efetuado através do presente link (Google Forms).
Resumos dos Oradores [Eng]
Os títulos e resumos dos oradores deste I Simpósio Online em Biometria Galaico-Portuguesa são aqui disponibilizados em inglês.
Amalia Jácome Pumar, Universidade da Coruña
Título: Testing the effect on the cure rate of multiple covariates using martingale difference correlation
Resumo: In survival analysis, there are situations in which not all subjects are susceptible to the final event. The population is divided into two parts: individuals who will experience the event (uncured) and those who will never suffer it (cured). Estimating the probability of experiencing the event is crucial in these situations. Methods are now well-developed to deal with this kind of data, generally known as cure model analysis. Mixture cure models allow for the estimation of the probability of cure and the survival function for uncured subjects. One of the main goals when working with these models is to determine if the covariates have any effect on the cure rate. We propose a new nonparametric test, using a new measure called martingale difference correlation. This extension of the distance correlation, measures the departure from conditional mean independence between a scalar response variable Y (the cure indicator in our case) and a vector predictor variable X (covariates). Moreover, the martingale difference correlation and its empirical counterpart inherit several desirable features from the distance correlation and the sample distance correlation. A new significance test for multiple covariates is proposed using martingale difference correlation and its extensions.
Rui Martins, Centro de Estatística e Aplicações da Universidade de Lisboa
Título: An overlooked prior distribution for proportions: the Logistic-Normal in dynamic football predictions
Resumo: The outcome of a football match—Home Win, Draw, or Away Win—can be modeled as a realization of a multinomial random variable with three mutually exclusive events. Traditional approaches often use the Dirichlet distribution as a prior for the multinomial proportions, primarily due to its conjugacy and simplicity. In this work, we propose an alternative: the Logistic-Normal distribution, a multivariate prior for proportions that has received limited attention in this context. This methodology was inspired by the question: Are women’s and men’s football leagues equally predictable? We apply our models to data from the main Portuguese women’s and men’s football leagues, covering seven full seasons (2016–2017 to 2022–2023). The analysis provides estimates of latent team-specific strengths, accounts for variability across seasons and match rounds, and explores the impact of home advantage. Additionally, we assess league competitiveness using appropriate metrics, offering fresh insights into the dynamics of these leagues.
María Gómez Rúa, Universidade de Vigo
Título: Additivity in biodiversity problems.
Resumo: We study biodiversity problems where a finite number of land sites have private owners. Land owners can use their land for either forests or for other uses. Since forests help to conserve biodiversity and are also less affected by climate change, they generate positive externalities over adjacent lands. A natural question is how to incentivize land owners to change their land use to forest use to maximize social benefit. In this paper, we study the property of additivity and characterize the rules that allow sharing the benefit of cooperation and satisfy this relevant property.
Rita Gaio, Universidade do Porto
Título: Testing the goodness-of-fit in generalized partially linear models
Resumo: We propose a novel projection-based test to check goodness of fit in generalized partially linear regression models using a residual-marked empirical process. The used residuals follow directly from the estimation process, which is valid whenever the means of the conditional response variable are not too small. The test is also adapted, and carried out, for a different structure of residuals, with no additional assumptions. The two tests are compared in terms of type I error control in simulated scenarios of a Poisson model with different samples sizes. This is a joint work with Rui Miranda and Wenceslao González-Manteiga.
- Amalia Jácome Pumar, Universidade da Coruña
- Rui Martins, Centro de Estatística e Aplicações da Universidade de Lisboa
- María Gómez Rúa, Universidade de Vigo
- Rita Gaio, Universidade do Porto
Informações e Registo
O Simpósio irá decorrer a 17/01/2025, sexta-feira, entre as 14:30h e as 16:30h (hora de Portugal; +1 hora de Galiza).
A participação é GRATUITA para todos, inclusive não-sócios da SPE/SGAPEIO. Contudo o registo é obrigatório para que a sessão possa ser disponibilizada.
O registo no Simpósio pode ser efetuado através do presente link (Google Forms).
Resumos dos Oradores [Eng]
Os títulos e resumos dos oradores deste I Simpósio Online em Biometria Galaico-Portuguesa são aqui disponibilizados em inglês.
Amalia Jácome Pumar, Universidade da Coruña
Título: Testing the effect on the cure rate of multiple covariates using martingale difference correlation
Resumo: In survival analysis, there are situations in which not all subjects are susceptible to the final event. The population is divided into two parts: individuals who will experience the event (uncured) and those who will never suffer it (cured). Estimating the probability of experiencing the event is crucial in these situations. Methods are now well-developed to deal with this kind of data, generally known as cure model analysis. Mixture cure models allow for the estimation of the probability of cure and the survival function for uncured subjects. One of the main goals when working with these models is to determine if the covariates have any effect on the cure rate. We propose a new nonparametric test, using a new measure called martingale difference correlation. This extension of the distance correlation, measures the departure from conditional mean independence between a scalar response variable Y (the cure indicator in our case) and a vector predictor variable X (covariates). Moreover, the martingale difference correlation and its empirical counterpart inherit several desirable features from the distance correlation and the sample distance correlation. A new significance test for multiple covariates is proposed using martingale difference correlation and its extensions.
Rui Martins, Centro de Estatística e Aplicações da Universidade de Lisboa
Título: An overlooked prior distribution for proportions: the Logistic-Normal in dynamic football predictions
Resumo: The outcome of a football match—Home Win, Draw, or Away Win—can be modeled as a realization of a multinomial random variable with three mutually exclusive events. Traditional approaches often use the Dirichlet distribution as a prior for the multinomial proportions, primarily due to its conjugacy and simplicity. In this work, we propose an alternative: the Logistic-Normal distribution, a multivariate prior for proportions that has received limited attention in this context. This methodology was inspired by the question: Are women’s and men’s football leagues equally predictable? We apply our models to data from the main Portuguese women’s and men’s football leagues, covering seven full seasons (2016–2017 to 2022–2023). The analysis provides estimates of latent team-specific strengths, accounts for variability across seasons and match rounds, and explores the impact of home advantage. Additionally, we assess league competitiveness using appropriate metrics, offering fresh insights into the dynamics of these leagues.
María Gómez Rúa, Universidade de Vigo
Título: Additivity in biodiversity problems.
Resumo: We study biodiversity problems where a finite number of land sites have private owners. Land owners can use their land for either forests or for other uses. Since forests help to conserve biodiversity and are also less affected by climate change, they generate positive externalities over adjacent lands. A natural question is how to incentivize land owners to change their land use to forest use to maximize social benefit. In this paper, we study the property of additivity and characterize the rules that allow sharing the benefit of cooperation and satisfy this relevant property.
Rita Gaio, Universidade do Porto
Título: Testing the goodness-of-fit in generalized partially linear models
Resumo: We propose a novel projection-based test to check goodness of fit in generalized partially linear regression models using a residual-marked empirical process. The used residuals follow directly from the estimation process, which is valid whenever the means of the conditional response variable are not too small. The test is also adapted, and carried out, for a different structure of residuals, with no additional assumptions. The two tests are compared in terms of type I error control in simulated scenarios of a Poisson model with different samples sizes. This is a joint work with Rui Miranda and Wenceslao González-Manteiga.
