SHSADREM

Censored extensions of distributional regression models will be developed by deriving the censored likelihood from the distributional regression model. In addition, new multivariate distributional regression models will be achieved through copulas, which in an integrated way, allows the combination of different marginal distributions with a multitude of dependence structures. This area is motivated by two important questions in the context of the Breast Cancer Screening Program (BCSP) and the Longitudinal Study on Personality Traits (LSPT) developed at the University of Coimbra.

By extending Rasch models to the structured additive regression class, the study proposes relaxing some of the underlying assumptions (e.g. logistic item characteristics curves with common slopes, unidimensionality and local dependence) necessary for them to produce a good performance in separating item difficulty and person ability estimates on an linear measurement scale. Bayesian inference, variable selection and regularization in distributional Rasch models will be developed to propose appropriate identification strategies and include distributional components representing an individual personal ability.

One important question arises in the context of identifying breast cancer risk factors in the investigation of the association with age at menarche and a woman's reproductive lifespan, i.e. the time between menarche and menopause. We apply multivariate distributional regression models to investigate these two important variables simultaneously and with a particular focus on the dependence structure and its relation to explanatory variables. In addition, age at menopause is naturally only available for women above a certain age. Thus, the screening data set contains registries where the age at menopause as well as a woman’s reproductive lifespan are censored. We will therefore develop censored extensions of distributional regression models that compensate for this censoring. As common practice in survival analysis, this can be achieved by deriving the censored likelihood from the distributional regression model. Nevertheless, such modifications will require modifications in the inferential principles developed for distributional regression.

A possible adverse effect of a screening program is the psychological impact on women when receiving an invitation to participate or when they are placed into surveillance and asked to come back 6-12 months later for a re-test. This adverse effect could deter women from attending their first or sub-sequent visits to the screening program, with the severe consequence of lessening the potential benefit of a screening program.  A new questionnaire was developed to measure the adverse psychological consequences of routine mammographic screening. The questionnaire was applied to a random sample of 1400 women who participated in the mammographic screening program of Portuguese Cancer League in Portugal. The results will contribute to the improvement in policies and practices in order to minimize possible adverse effects which may undermine the potential benefits of a screening program.

In this study, we focus on the concepts of citizenship, good citizenship, and Citizenship Education (CE), its perceptions, and the pertinence of these concepts. The main concern is that the overvaluation given to education for the labor market (vocational programs) and the absence of social education could lead to a less just, less democratic and more conflicted society. Thus, there is a clear need to strike a balance between the skills for the labor market and social skills in the school curricula, teaching practices and, in particular, teacher training. For this purpose, a questionnaire was constructed. This questionnaire was given to a random sample of Portuguese upper secondary teachers in public schools with academic and/or vocational programs, taking into consideration school characteristics (e.g. dimension of the school).  Results are expected to contribute to Portuguese educational policies and practices, aiming at strengthening citizenship, in harmony with the economic dimension in education.

The longitudinal data collected over the last 20 years in Coimbra will allow us to address some complex questions such as how people feel, how they think, and what they want. By addressing individual differences and their evolution over time, we hope to contribute to a clearer picture of what leads to individual behavior. These complex relationships will be addressed through distributional regression models based on parametric assumptions. This approach will allow for multivariate distributions to be integrated effortlessly as long as an appropriate multivariate distribution for the response vector is available. Thus, flexible multivariate specifications will be derived, for example, based on copulas, that will combine various specifications for the marginal distribution with different types of dependence structures in a modular manner.