Affichage des résultats 61 à 80 sur 1312 au total
Papers evaluating measures of explained variation, or similar indices, invariably use independence from censoring as the most important criterion. And they invariably end up suggesting that some measures meet this criterion, and some don't, leading to a conclusion that the first are better than the second. As a consequence, users are offered measures that cannot be used with time-dependant covariates and effects, not to mention extensions to repeated events or multi state models. We explain in this paper that the above mentioned criterion is of no use in studying such measures, since it simply favours those that make an implicit assumption of a model being valid everywhere. Measures not making such an assumption are disqualified, even though they are better in every other respect. We show that if these, allegedly inferior, measures are allowed to make the same assumption, they are easily corrected to satisfy the `independent-from-censoring' criterion. Even better, it is enough to make such an assumption only for the times greater than the last observed failure time $tau$. Which, in contrast with the `preferred' measures, makes it possible to use all the modelling flexibility up-to $tau$, and assume whatever one wants after $tau$. As a consequence, we claim that measures being proffered as better in the existing reviews, are exactly those that are inferior
-
-
-
Le succès de transmission d'une maladie infectieuse repose entre autres sur les interactions entre des individus susceptibles et des personnes contagieuses. L'acquisition d'une infection par un patient dans un établissement de santé n'échappe pas à ce schéma. En effet, les autres patients, les professionnels de santé, les visiteurs sont des sources potentielles d'infection pour les patients. La quantification de la contribution de ces différentes expositions sur le risque d'infection donnerait des informations utiles pour la prévention et le contrôle des épidémies hospitalières. Dans une première partie, à travers différents modèles statistiques, nous étudierons comment l'exposition à différentes sources modifie le risque d'acquisition d'une infection grippale parmi les patients. Dans une seconde partie, nous discuterons de la mesure de l'exposition à l'aide d'une mesure électronique des contacts, et du niveau de détails utile à introduire dans les modèles mathématiques de prédiction. Les résultats ont des retombées en termes d'épidémiologie hospitalière et de modélisation.
-
The early developmental period of an organism is a sensitive period where organisational and activational effects occur and thus, small disturbances have important and long-lasting consequences. Parents can modify, to some extent, the abiotic and biotic conditions experienced by their offspring, during both the prenatal and postnatal stages through differential behavioural and/or physiological input. In this seminar, I will focus on prenatal effects, such as litter-sex composition and maternal programming for the future environment in yellow-bellied marmots (Marmota flaviventris), a ground-dwelling sciurid rodent.
We propose a new approach, along with refinements, based on L1 penalties and aimed at jointly estimating several related regression models. It is especially useful in epidemiology and clinical research when data come from several strata of a population. The main interest of our approach is that it can be rewritten as a weighted lasso on a simple transformation of the original data set. In particular, it does not need new dedicated algorithms and is ready to implement under a variety of regression models, e.g. linear or logistic models using standard R packages. Moreover, asymptotic oracle properties are derived along with preliminary non-asymptotic results, suggesting good theoretical properties. Our approach is further compared with state-of-the-art competitors under various settings on synthetic data: these empirical results confirm that our approach performs at least similarly to its competitors. As a final illustration, an analysis of road safety data is provided.
We consider the issue of estimating causal effects in a dynamic approach based on a multivariate stochastic process representation, which may be called the "stochastic system approach". Conditional and marginal effects can be defined. We focus on the issue of the horizon on which causal influence must be studied, in particular in ageing studies. In ageing studies, one of the most important events that we have to consider is "death". This is why the illness-death model is important in such studies. But "death" is not an event which is on the same footing as other events that can happen to subjects. Even if the vital status is part of the state, it has a very special meaning, in that all the other components of the state are defined only for a living subject. The consequence is that causal influences must be defined on a maximum horizon which is the time of death. We do not say that death has an influence on the other components of the state, but that these other components are not defined after death. For instance if we are interested in dementia, the state can be represented by a bivariate counting process counting dementia and death. However dementia is defined only for a living subject: after death the subject does not exist anymore and cannot be qualified as demented or not demented. When we investigate the causal influence of a factor, we should first look at its causal influence on death, then on its influences on other processes. Cases where a value of a modifiable factor can be preferred to another one will be given. Thus, the stochastic system approach helps clarifying the important issue of assessing causal effects in ageing studies Contact : Mme Mariethé CHAUMEIL Inscription gratuite mais obligatoire (commande sandwichs avant le lundi 16 MARS 2015 Courriel : mariethe.chaumeil@chu-lyon.fr
Vibrios have been associated with successive mortality outbreaks ofCrassostrea gigas) in France that have resulted in losses up to 100% of production. Given the near monoculture of C. gigas in Europe, there is an urgent need to understand the epidemiology of these outbreaks, particularly the role of Vibrio in the diseases.The study of the Vibrios distribution on fine phylogenetic and spatial scales has demonstrated that vibrios coexisting in the water column can be divided into closely related populations, which pursue different lifestyles i.e. ecological population (Hunt et al., 2008). However, a link between ecological populations and pathogenicity has not been demonstrated, and it is unclear whether pathogenicity is a trait primarily linked to clones or to populations comprising a large number of distinct genotypes.In the presentVibrio populations in an intensive oyster cultivation area. We demonstrate that Vibrio populations do not assemble neutrally in oysters from water column populations i.e. specific genotypes colonize the oysters. Combining experimental ecology, high throughput infection assay and genome sequencing, we showed that the onset of disease in oysters is associated with progressive replacement of diverse, benign colonizers by members of a phylogenetically coherent virulent population together with quorum sensing pheromone producers. Analyses of oyster mortality following experimental infection suggest that disease onset can be facilitated by the presence of non-virulent strains. Oyster disease may thus represent a new form of polymicrobial disease, in which non-pathogenic strains contribute to increased mortality.Hunt DE, et al. (2008) Resource partitioning and sympatric differentiation among closely related bacterioplankton. Science 320(5879):1081-1085.Lemire A, Goudenège D, Versigny T, Petton B, Calteau A, Labreuche Y, Le Roux F. (2014) Populations, not clones, are the unit of vibrio pathogenesis in naturally infected oysters. ISME J. Dec 9. doi: 10.1038/ismej.2014.233
-
-
Among biologist as well as linguists, it is now widely accepted that there are many striking parallels between the evolution of life forms and the history of languages. Starting from the rise of language studies as a scientific discipline in the early 19th century, up to today's recent "quantitative turn" in historical linguistics, scholars from both disciplines have repeatedly pointed to similarities between the respective research objects in biology and linguistics. During the last two decades, this has lead to a new school of ''quantitative historical linguistics''. Based on the key assumption that the characteristic processes of language change and biological evolution are so similar that the methods designed for one discipline may also be used in the other one, methods which were originally designed to study biological evolution (methods for phylogenetic reconstruction, sequence alignment, or biological network analysis) have now repeatedly been applied to linguistic data.Unfortunately, not all analogies which have been made between evolutionary processes in linguistics and biology reflect true similarities in the processes. Striking differences between the research objects of both disciplines are often ignored. In the talk, I will review proposed similarities between evolutionary processes in the two disciplines and discuss their methodological implications.
A recent paper in Science (January 2015) claimed that the majority, 64% to be precise, of cancers is due to bad luck, so non-preventable. The message was spread quickly through media, including serious ones like BBC (and also radio Slovenia, if I may add). And while the paper has been criticized by many, the authors seem to stick to the original message.In this talk I'll give my view of the paper and try to defend a counter message that their analysis gives them absolutely no grounds to make such a claim. The arguments that I'll present have, to my knowledge, not appeared in published reactions to the paper. Obviously, if I am right, they are wrong, and vice versa. If I was Bayesian, I would, at present, give a 0.99 prior probability to the first option. Either way, it should be fun.
-
-
-
La dispersion est un processus central pour le maintien de la biodiversité dans les paysages. Cependant les patterns de dispersion et leurs déterminants sont encore mal documentés, en particulier chez les organismes du sol. En conséquence, une grande partie des déterminants de la biodiversité des sols n'est toujours pas comprise.Nous exposons les résultats d'un projet ANR de quatre ans sur la dispersion des vers de terre, un organisme clé dans le maintien de nombreux services écosystémiques. Avec une combinaison d'approches (1) de terrain multi-échelle, (2) de (méta) communauté, (3) de génétique du paysage, et (4) d'expérience en laboratoire, nous avons identifié des mécanismes centraux qui permettent aux vers de terre de disperser dans un agro-écosystème de manière adaptée, et nous montrons l'importance de la dispersion pour le maintien de la diversité spécifique et génétique.Nos résultats suggèrent également que la dispersion passive joue un rôle sous estimé dans la dynamique spatiale de ces organismes.
-
We are exploring techniques using single cell genomic sequencing in human beings for the purpose of defining phylogenetic trees to explain the relationship of cells within defined human tissues. I will discuss how we used single cell exome and whole genome sequencing to identify patient and donor adipocytes in human white adipose tissue biopsies taken from bone marrow recipients and discuss future projects using similar methods to explore cellular identities in human tissue samples.