Dimitris Fouskakis | National Technical University of Athens (original) (raw)

Papers by Dimitris Fouskakis

arXiv (Cornell University), Feb 13, 2020

One of the main approaches used to construct prior distributions for objective Bayes methods is t... more One of the main approaches used to construct prior distributions for objective Bayes methods is the concept of random imaginary observations. Under this setup, the expected-posterior prior (EPP) offers several advantages, among which it has a nice and simple interpretation and provides an effective way to establish compatibility of priors among models. In this paper, we study the power-expected-posterior prior as a generalization to the EPP in objective Bayesian model selection under normal linear models. We prove that it can be represented as a mixture of g-prior, like a wide range of prior distributions under normal linear models, and thus posterior distributions and Bayes factors are derived in closed form, keeping therefore its computational tractability. Following this result, we can naturally prove that desiderata (criteria for objective Bayesian model comparison) hold for the PEP prior. Comparisons with other mixtures of g-prior are made and results are presented in simulated and real-life datasets.

Biometrika, 2007

In this paper we present an extension of population-based Markov chain Monte Carlo (MCMC) to the ... more In this paper we present an extension of population-based Markov chain Monte Carlo (MCMC) to the trans-dimensional case. One of the main challenges in MCMC-based inference is that of simulating from high and trans-dimensional target measures. In such cases, MCMC methods may not adequately traverse the support of the target; the simulation results will be unreliable. We develop population methods to deal with such problems, and give a result proving the uniform ergodicity of these population algorithms, under mild assumptions. This result is used to demonstrate the superiority, in terms of convergence rate, of a population transition kernel over a reversible jump sampler for a Bayesian variable selection problem. We also give an example of a population algorithm for a Bayesian multivariate mixture model with an unknown number of components. This is applied to gene expression data of 1000 data points in six dimensions and it is demonstrated that our algorithm out performs some competing Markov chain samplers.

Journal of statistical theory and practice, Aug 31, 2022

An Excel file that contains all the responses to the inventory and additional information for eac... more An Excel file that contains all the responses to the inventory and additional information for each responder (see Dedos S.G. & Fousakis D. manuscript for details)<br><br>

Data Record (Data Citation 1) The Greek version of the Approaches to Study Skills Inventory for S... more Data Record (Data Citation 1) The Greek version of the Approaches to Study Skills Inventory for Students

This is an SPSS output file, in .htm format, that contains all the dataset's validation resul... more This is an SPSS output file, in .htm format, that contains all the dataset's validation results<br>

Factors influencing fetal and childhood growth may affect a person’s risk of developing schizophr... more Factors influencing fetal and childhood growth may affect a person’s risk of developing schizophrenia. Associations of size at birth and body size in young adulthood with schizophrenia and other nonaffective psychoses were assessed in a cohort of 334,577 Swedish male conscripts born in 1973–1980 for whom linked birth, census, hospital admission, and adult height and weight data were available. Complete data on all study variables were available for 246,655 subjects. Over a mean 3.4-year follow-up beginning at age 18 years, 80 subjects developed schizophrenia and 124 developed other nonaffective psychoses. A reverse J-shaped association was found between gestation-adjusted birth weight and schizophrenia. The hazard ratios were 7.03 (95 % confidence interval: 1.59, 31.10) for males of low birth weight (<2.5 kg) and 3.37 (95 % confidence interval: 1.68, 6.74) for those of high birth weight (>4.0 kg). Birth weight was not strongly related to other nonaffective psychoses. Taller ma...

Scientific Data, 2021

There exists a vast amount of research on how students, in higher education, approach their study... more There exists a vast amount of research on how students, in higher education, approach their studying and learning. Such research resulted in a multitude of questionnaires and tools to capture the way students perform in higher education institutions. One of these tools is the Approaches to Study Skills Inventory for Students (ASSIST) that was developed in the ’80 s and ’90 s. This inventory broadly classifies students, as approaching their study, in a deep, a strategic and/or a surface manner. Although it has gone through rigorous validation in many studies, there exist no publicly available dataset of the results of these studies and so the raw datasets cannot be re-used or integrated with other similar datasets. Here, we report and make publicly available the raw data of an 8-year longitudinal survey using this inventory in a cohort study of 1181 students from a department of a higher education institution. We validated our dataset using reliability analyses that confirmed, and co...

Landscape Ecology, 2008

Pieter Jan’t Hoen A.E. Eiben Krzysztof Krawiec Hussain Abbass Aniko Ekart Mario Köppen Chang Wook... more Pieter Jan’t Hoen A.E. Eiben Krzysztof Krawiec Hussain Abbass Aniko Ekart Mario Köppen Chang Wook Ahn Michael Emmerich Dario Landa-Silva Juan Manuel Ahuactzin Andries Engelbrecht Pedro Larrañaga Lourdes Araujo Anton Eremeev Loo Hay Lee Dirk Arnold Francisco Fernandez Per Kristian Lehre Anne Auger Dario Floreano Rhydian Lewis Victor Ayala Gianluigi Folino Xiaodong Li Jaume Bacardit Cyril Fonlupt Lucian Lita Lucia Ballerini Dimitris Fouskakis Daniele Loiacono Pedro Ballester Jeremy Frank Jose Santamaria Lopez Raul Banos Marcus Gallagher Manuel Lopez-Ibanez Peter Bentley Luca M. Gambardella Jean Louchet Ester Bernadó Mansilla Jose Gamez Jose Antonio Lozano Tim Blackwell Yong Gao Sean Luke Isabelle Bloch Mario Giacobini Evelyne Lutton Leonardo Bocchi Jose Luis Gordillo Juan J. Merelo Lashon Booker Jens Gottlieb Silja Meyer-Nieberg Yosef Borenstein Jörn Grahl Julian Miller Nicolas Bredèche Steven Gustafson Alfonsas Misevicius Larry Bull Riad Hammoud Boris Mitavskiy Martin Butz Nikolaus H...

Statistica Neerlandica, 2020

Graphical models are used for expressing conditional independence relationships among variables b... more Graphical models are used for expressing conditional independence relationships among variables by the means of graphs, whose structure is typically unknown and must be inferred by the data at hand. We propose a theoretically sound Objective Bayes procedure for graphical model selection. Our method is based on the Expected-Posterior Prior and on the Power-Expected-Posterior Prior. We use as input of the proposed methodology a default improper prior and suggest computationally efficient approximations of Bayes factors and posterior odds. In a variety of simulated scenarios with varying number of nodes and sample sizes, we show that our method is highly competitive with, or better than, current benchmarks. We also discuss an application to protein-signaling data, which wieldy confirms existing results in the scientific literature.

METRON, 2020

The aim of the paper is to estimate the posterior mean values and analyze the posterior variation... more The aim of the paper is to estimate the posterior mean values and analyze the posterior variation in students' prioritization of teaching quality components within a 10-year frame. The results are based on longitudinal data gathered among Greek university students during the period of the national economic crisis in Greece spanned from 2009 to 2018. The analysis consists of fitting a Bayesian hierarchical beta regression model with a Dirichlet prior on the model coefficients that correspond to twenty quality attribute measures. Using this natural way to implement the usual constraints, the model coefficients can be interpreted as weights and thus they measure the relative importance that the students give to the different attributes. By estimating the posterior means and positioning measures of all consecutive sampling instances and summarizing posterior distributions of the differences between consecutive periods in the model weights, the study identifies and evaluates the major changes and patterns in students' perception of academic quality over the ten-year sampling period.

Statistical Theory and Related Fields, 2020

The power-expected-posterior prior is used in this paper for comparing nested linear models. The ... more The power-expected-posterior prior is used in this paper for comparing nested linear models. The asymptotic behaviour of the method is investigated for different values of the power parameter of the prior. Focus is given on the consistency of the Bayes factor of comparing the full model M p versus a generic submodel M. In each case, we allow the true generating model to be either M p or M and we keep the dimension of M fixed, while the dimension of M p can be either fixed or (grow as) O(n), with n denoting the sample size.

Tobacco, smoking control and health education, 2018

Journal of Applied Statistics, 2015

The problem motivating the paper is the quantification of students preferences regarding teaching... more The problem motivating the paper is the quantification of students preferences regarding teaching/coursework quality, under certain numerical restrictions, in order to build a model for identifying, assessing and monitoring the major components of the overall academic quality. After reviewing the strengths and limitations of conjoint analysis and of the random coefficient regression model used in similar problems in the past, we propose a Bayesian beta regression model with a Dirichlet prior on the model coefficients. This approach not only allows for the incorporation of informative prior when it is available but also provides user friendly interfaces and direct probability interpretations for all quantities. Furthermore, it is a natural way to implement the usual constraints for the model weights/coefficients. This model was applied to data collected in 2009 and 2013 from undergraduate students in Panteion University, Athens, Greece and besides the construction of an instrument for the assessment and monitoring of teaching quality, it gave some input for a preliminary discussion on the association of the differences in students preferences between the two time periods with the current Greek economic and financial crisis.

Anticancer research

Several factors are currently employed for prognosis assessment and treatment determination in br... more Several factors are currently employed for prognosis assessment and treatment determination in breast cancer. An array of molecular parameters, such as p53, Her2-neu (c-erbB 2) and Cathepsin-D, are also examined to improve clinical patient management. We have conducted a statistically powerful study of the prognostic value of conventional factors and of the investigational factors p53, Her2-neu and Cathepsin-D in patients with invasive breast carcinoma, in order to compare their significance. Our analysis was extended to determine the associations of p53 and Her2-neu with risk of death and relapse among patients with and without lymph node metastases. In a set of 125 primary breast tumors, p53 and Her2-neu expression were immunohistochemically evaluated. Cathepsin-D, estrogen and progesterone receptor concentrations were determined in cytosols by a standard immunoradiometric assay. Over a mean of 62 months, 49 patients (39%) had a relapse and 29 patients (23%) died. Overexpression o...

Psychological Medicine, 2003

Background. Although urban place of birth has been identified as a risk factor for schizophrenia,... more Background. Although urban place of birth has been identified as a risk factor for schizophrenia, the extent to which this association is mediated by socially patterned risk factors such as obstetric complications and childhood socio-economic position is unclear. The diagnostic specificity of the association within the clinical psychotic syndromes is also unclear.Method. A population cohort of 696025 males and females, born in Sweden between 1973 and 1980 and with linked birth and socio-economic data was followed up from age 16 for up to 9·8 years. Hospitalized cases of schizophrenia and other non-affective psychosis were identified from the Swedish Inpatient Discharge Register. We examined associations of these disorders with a three-level measure of urbanicity of birthplace before and after controlling for measures of foetal nutrition, obstetric complications and level of maternal education.Results. Urban compared to rural birthplace was associated both with increased risk of adul...

PLoS ONE, 2012

Background: The traditional Serfling-type approach for influenza-like illness surveillance requir... more Background: The traditional Serfling-type approach for influenza-like illness surveillance requires long historical time-series. We retrospectively evaluated the use of recent, short, historical time-series for recognizing the onset of community outbreaks of respiratory tract infections (RTIs). Methods: The data used referred to the proportion of diagnoses for upper or lower RTIs to total diagnoses for house-call visits, performed by a private network of medical specialists (SOS Doctors) in the metropolitan area of Athens, Greece, between January 01, 2000 and October 12, 2008. The reference standard classification of the observations was obtained by generating epidemic thresholds after analyzing the full 9-year period. We evaluated two different alert generating methods [simple regression and cumulative sum (CUSUM), respectively], under a range of input parameters, using data for the previous running 4-6 week period. These methods were applied if the previous weeks contained non-aberrant observations. Results: We found that the CUSUM model with a specific set of parameters performed marginally better than simple regression for both groups. The best results (sensitivity, specificity) for simple regression and CUSUM models for upper RTIs were (1.00, 0.82) and (0.94, 0.93) respectively. Corresponding results for lower RTIs were (1.00, 0.80) and (0.93, 0.91) respectively. Conclusions: Short-term data for house-call visits can be used rather reliably to identify respiratory tract outbreaks in the community using simple regression and CUSUM methods. Such surveillance models could be particularly useful when a large historical database is either unavailable or inaccurate and, thus, traditional methods are not optimal.

International Journal of Gynecology & Obstetrics, 2011

To determine the maternal and perinatal outcome for different types of placenta previa (PP). Meth... more To determine the maternal and perinatal outcome for different types of placenta previa (PP). Methods: A retrospective review of 132 singleton pregnancies with PP. Outcome measures, including the incidence of obstetric hysterectomy, the neonatal Apgar score, and the neonatal weight, were evaluated by logistic regression analysis. Results: The incidence of PP was 1.0%. Of the women with PP, 51.5% had complete PP, 20.5% had incomplete PP, 5.3% had marginal PP, and 22.7% had a low-lying placenta. Most (93.9%) women were delivered by cesarean delivery. In total, 19.7% women underwent obstetric hysterectomy; of these, 92.3% had complete PP. Mothers with 2 or more previous cesarean deliveries had an increased risk for obstetric hysterectomy (P b 0.01). The gestational age at delivery was a significant linear predictor of the 5minute Apgar score. Mothers with incomplete PP delivered neonates with lower Apgar scores than did mothers with complete PP. Conclusion: A history of multiple cesarean deliveries increased the risk for obstetric hysterectomy in women with PP. The type of PP had no effect on maternal and neonatal outcome, with exception of the fact that neonates in the incomplete PP group had lower Apgar scores than neonates in the complete PP group.