Suboptimal Multistage Nonparametric Hypotheses Test

2000 Mathematics Subject Classification: 62L10, 62L15.

Nonparametric versus Parametric Statistical Approaches for Genetic Anticipation: The Pancreatic Cancer Case

2000 Mathematics Subject Classi cation: 62N01, 62N05, 62P10, 92D10, 92D30.

Magnetoencephalography in epilepsy

Magnetoencephalographic (MEG) signals, like electroencephalographic (EEG) measures, are the direct extracranial manifestations of neuronal activation. The two techniques can detect time-varying changes in electromagnetic activity with a sub-millisecond time resolution. Extra-cranial electromagnetic measures are the cornerstone of the non-invasive diagnostic armamentarium in patients with epilepsy. Their extremely high temporal resolution – comparable to intracranial recordings – is the basis for a precise definition of onset and propagation of ictal and interictal abnormalities. Given the cost of the infrastructure and equipment, MEG has yet to develop into a routinely applicable diagnostic tool in clinical settings. However, in recent years, an increasing number of patients with epilepsy have been investigated – usually in the context of presurgical evaluation of refractory epilepsies – and initial encouraging results have been reported. We will briefly review the principles and the technology behind MEG and its contribution in the diagnostic work-up of patients with epilepsy.

Optimising QoE distribution for video applications through LTE-WiFi interworking

Mobile WiFi devices are becoming increasingly popular in non-seamless and user-controlled mobile traffic offloading alongside the standard WiFi hotspots. Unlike the operator-controlled hotspots, a mobile WiFi device relies on the capacity of the macro-cell for the data rate allocated to it. This type of devices can help offloading data traffic from the macro-cell base station and serve the end users within a closer range, but will change the pattern of resource distributions operated by the base station. We propose a resource allocation scheme that aims to optimize user quality of experience (QoE) when accessing video services in the environment where traffic offloading is taking place through interworking between a mobile communication system and low range wireless LANs. In this scheme, a rate redistribution algorithm is derived to perform scheduling which is controlled by a no-reference quality assessment metric in order to achieve the desired trade-offs between efficiency and fairness. We show the performance of this algorithm in terms of the distribution of the allocated data rates throughout the macro-cell investigated and the service coverage offered by the WiFi access point.

Enhanced connectivity and QoE through LTE-WiFi interworking in Smart Cities

Mobile communication and networking infrastructures play an important role in the development of smart cities, to support real-time information exchange and management required in modern urbanization. Mobile WiFi devices that help offloading data traffic from the macro-cell base station and serve the end users within a closer range can significantly improve the connectivity of wireless communications between essential components including infrastructural and human devices in a city. However, this offloading function through interworking between LTE and WiFi systems will change the pattern of resource distributions operated by the base station. In this paper, a resource allocation scheme is proposed to ensure stable service coverage and end-user quality of experience (QoE) when offloading takes place in a macro-cell environment. In this scheme, a rate redistribution algorithm is derived to form a parametric scheduler to meet the required levels of efficiency and fairness, guided by a no-reference quality assessment metric. We show that the performance of resource allocation can be regulated by this scheduler without affecting the service coverage offered by the WLAN access point. The performances of different interworking scenarios and macro-cell scheduling policies are also compared.

Impaired brainstem and thalamic high-frequency oscillatory EEG activity in migraine between attacks

INTRODUCTION: We investigated whether interictal thalamic dysfunction in migraine without aura (MO) patients is a primary determinant or the expression of its functional disconnection from proximal or distal areas along the somatosensory pathway. METHODS: Twenty MO patients and twenty healthy volunteers (HVs) underwent an electroencephalographic (EEG) recording during electrical stimulation of the median nerve at the wrist. We used the functional source separation algorithm to extract four functionally constrained nodes (brainstem, thalamus, primary sensory radial, and primary sensory motor tangential parietal sources) along the somatosensory pathway. Two digital filters (1-400 Hz and 450-750 Hz) were applied in order to extract low- (LFO) and high- frequency (HFO) oscillatory activity from the broadband signal. RESULTS: Compared to HVs, patients presented significantly lower brainstem (BS) and thalamic (Th) HFO activation bilaterally. No difference between the two cortical HFO as well as in LFO peak activations between the two groups was seen. The age of onset of the headache was positively correlated with HFO power in the right brainstem and thalamus. CONCLUSIONS: This study provides evidence for complex dysfunction of brainstem and thalamocortical networks under the control of genetic factors that might act by modulating the severity of migraine phenotype.

Human disturbance and forest diversity in the Tansa Valley, India

We assessed the diversity of woody plants at 15 forested sites in the Tansa Valley of Thane District, in Maharashtra, India. The fewest species (11) were seen at a degraded mangrove site near the river mouth, and the greatest number (150) in the rich semi-evergreen forest on Tungar Hill. For all sites there were 141 tree, 25 shrub and 15 liana species, a total of 181 species. Excluding the mangrove site, which had no species in common with the other 14 sites, we analyzed the species distributions in detail. 2 These sites ranged in area from 4 to 30 km each, had woody floras of 89 6 6 species, and varied in intensity of human impact. Despite a history of exploitation and substantial reduction in biomass from firewood collecting, set fires and illicit tree felling, considerable plant diversity remains in the area.We found a modest increase in species richness in transects away from two villages. We observed the exploitation of the forest by the principal users, primarily of the Warli Tribe. They exploited a wide variety of forest resources (92 species), for medicines, foods, construction materials, household goods, manure and other purposes. They collected 15 items for sale. By far the single most important item collected was firewood, which dramatically reduced forest biomass within 2 km of villages. The species distributions in these forest remnants are strongly nested, mostly due to varying degrees of disturbance at individual sites. The high species diversity on Tungar Hill is most likely a relict of the earlier character of forests throughout much of the valley. It merits the highest priorities for preservation, as a refuge for Western Ghat species at the northern limits of their distributions.

Um estudo sobre o perfil sexual e reprodutivo das adolescentes de um serviço de referência no município de Natal/RN

Adolescence is seen as a phase of life marked by a series of physical and behavioral changes, which leads to certain risk situations, such as early sexual activity, pregnancy and the occurrence of sexually transmitted diseases. Based on this statement, this research aimed generally: meet the sexual and reproductive adolescents of a reference service in Natal profile. Specifically describe the socioeconomic, sexual and reproductive individual and family characteristics of the population in question, but also verify possible association between pregnancy and age of onset of sexual activity with the socioeconomic aspects. The sectional study conducted in the database from the records of 463 adolescents seeking care in a referral center in Sexual and Reproductive Health in the period March 2011 to June 2012. The data collected were subjected to analysis by Excel 2007 and Statistical Package for Social Sciences (SPSS) 17.0. For data analysis, descriptive statistics, with absolute numbers and percentages was used, and its presentation by means of distribution and frequency tables. Results showed that the parents of these adolescents (65.7% - father; 57.8% - mother), had primary education; family income less than two minimum wages (66.2%); any type of chemical dependency in the family (33.5%) and presence of domestic violence (20.6%). A higher probability of pregnancy was verified when the mother had only primary education (26.3%), persons other than the father or mother contributed to family income (33.3%) and (26%) when there was substance abuse in the family early onset of sexual activity was observed when the mother had primary education (57.3%), persons other than the father or mother contributed to family income (63.1%) and the use of drugs / alcohol as a problem family (67.6%). It was concluded that such vulnerabilities appear as a reflection of the low social status of these young people, aggravated by emotional inheritance that is offered to every individual from birth. These findings may perhaps, support public health surveillance policies for adolescents in several areas

Modelagem estatística e atribuições dos eventos de precipitação extrema na Amazônia brasileira

Intense precipitation events (IPE) have been causing great social and economic losses in the affected regions. In the Amazon, these events can have serious impacts, primarily for populations living on the margins of its countless rivers, because when water levels are elevated, floods and/or inundations are generally observed. Thus, the main objective of this research is to study IPE, through Extreme Value Theory (EVT), to estimate return periods of these events and identify regions of the Brazilian Amazon where IPE have the largest values. The study was performed using daily rainfall data of the hydrometeorological network managed by the National Water Agency (Agência Nacional de Água) and the Meteorological Data Bank for Education and Research (Banco de Dados Meteorológicos para Ensino e Pesquisa) of the National Institute of Meteorology (Instituto Nacional de Meteorologia), covering the period 1983-2012. First, homogeneous rainfall regions were determined through cluster analysis, using the hierarchical agglomerative Ward method. Then synthetic series to represent the homogeneous regions were created. Next EVT, was applied in these series, through Generalized Extreme Value (GEV) and the Generalized Pareto Distribution (GPD). The goodness of fit of these distributions were evaluated by the application of the Kolmogorov-Smirnov test, which compares the cumulated empirical distributions with the theoretical ones. Finally, the composition technique was used to characterize the prevailing atmospheric patterns for the occurrence of IPE. The results suggest that the Brazilian Amazon has six pluvial homogeneous regions. It is expected more severe IPE to occur in the south and in the Amazon coast. More intense rainfall events are expected during the rainy or transitions seasons of each sub-region, with total daily precipitation of 146.1, 143.1 and 109.4 mm (GEV) and 201.6, 209.5 and 152.4 mm (GPD), at least once year, in the south, in the coast and in the northwest of the Brazilian Amazon, respectively. For the south Amazonia, the composition analysis revealed that IPE are associated with the configuration and formation of the South Atlantic Convergence Zone. Along the coast, intense precipitation events are associated with mesoscale systems, such Squall Lines. In Northwest Amazonia IPE are apparently associated with the Intertropical Convergence Zone and/or local convection.

Caracterização do ritmo circadiano de atividade e repouso em saguis idosos (Callithrix Jacchus)

Advanced age may become a limiting factor for the maintenance of rhythms in organisms, reducing the capacity of generation and synchronization of biological rhythms. In this study, the influence of aging on the expression of endogenous periodicity and synchronization (photic and social) of the circadian activity rhythm (CAR) was evaluated in a diurnal primate, the marmoset (Callithrix jacchus). This study had two approaches: one with longitudinal design, performed with a male marmoset in two different phases: adult (three years) and older (9 y.o.) (study 1) and the second, a transversal approach, with 6 old (♂: 9.7 ± 2.0 y.o.) and 11 adults animals (♂: 4.2 ± 0.8 y.o.) (study 2). The evaluation of the photic synchronization involved two conditions in LD (natural and artificial illuminations). In study 1, the animal was subjected to the following stages: LD (12:12 ~ 350: ~ 2 lx), LL (~ 350 lx) and LD resynchronization. In the second study, the animals were initially evaluated in natural LD, and then the same sequence stages of study 1. During the LL stage in study 2, the vocalizations of conspecifics kept in natural LD on the outside of the colony were considered temporal cue to the social synchronization. The record of the activity was performed automatically at intervals of five minutes through infrared sensor and actimeters, in studies 1 and 2, respectively. In general, the aged showed a more fragmented activity pattern (> IV < H and > PSD, ANOVA, p < 0.05), lower levels of activity (ANOVA, p < 0.05) and shorter duration of active phase (ANOVA, p < 0.05) in LD conditions, when compared to adults. In natural LD, the aged presented phase delay pronounced for onset and offset of active phase (ANOVA, p < 0.05), while the adults had the active phase more adjusted to light phase. Under artificial LD, there was phase advance and greater adjustment of onset and offset of activity in relation to the LD in the aged (ANOVA, p < 0.05). In LL, there was a positive correlation between age and the endogenous period () in the first 20 days (Spearman correlation, p < 0.05), with prolonged  held in two aged animals. In this condition, most adults showed free-running period of the circadian activity rhythm with  < 24 h for the first 30 days and later on relative coordination mediated by auditory cues. In study 2, the cross-correlation analysis between the activity profiles of the animals in LL with control animals kept under natural LD, found that there was less social synchronization in the aged. With the resubmission to the LD, the resynchronization rate was slower in the aged (t-test; p < 0.05) and in just one aged animal there was a loss of resynchronization capability. According to the data set, it is suggested that the aging in marmosets may be related to: 1) lower amplitude and greater fragmentation of the activity, accompanied to phase delay with extension of period, caused by changes in a photic input, in the generation and behavioral expression of the CAR; 2) lower capacity of the circadian activity rhythm to photic synchronization, that can become more robust in artificial lighting conditions, possibly due to the higher light intensities at the beginning of the active phase due to the abrupt transitions between the light and dark phases; and 3) smaller capacity of non-photic synchronization for auditory cues from conspecifics, possibly due to reducing sensory inputs and responsiveness of the circadian oscillators to auditory cues, what can make the aged marmoset most vulnerable, as these social cues may act as an important supporting factor for the photic synchronization.

Caracterização do ritmo circadiano de atividade e repouso em saguis idosos (Callithrix Jacchus)

Advanced age may become a limiting factor for the maintenance of rhythms in organisms, reducing the capacity of generation and synchronization of biological rhythms. In this study, the influence of aging on the expression of endogenous periodicity and synchronization (photic and social) of the circadian activity rhythm (CAR) was evaluated in a diurnal primate, the marmoset (Callithrix jacchus). This study had two approaches: one with longitudinal design, performed with a male marmoset in two different phases: adult (three years) and older (9 y.o.) (study 1) and the second, a transversal approach, with 6 old (♂: 9.7 ± 2.0 y.o.) and 11 adults animals (♂: 4.2 ± 0.8 y.o.) (study 2). The evaluation of the photic synchronization involved two conditions in LD (natural and artificial illuminations). In study 1, the animal was subjected to the following stages: LD (12:12 ~ 350: ~ 2 lx), LL (~ 350 lx) and LD resynchronization. In the second study, the animals were initially evaluated in natural LD, and then the same sequence stages of study 1. During the LL stage in study 2, the vocalizations of conspecifics kept in natural LD on the outside of the colony were considered temporal cue to the social synchronization. The record of the activity was performed automatically at intervals of five minutes through infrared sensor and actimeters, in studies 1 and 2, respectively. In general, the aged showed a more fragmented activity pattern (> IV < H and > PSD, ANOVA, p < 0.05), lower levels of activity (ANOVA, p < 0.05) and shorter duration of active phase (ANOVA, p < 0.05) in LD conditions, when compared to adults. In natural LD, the aged presented phase delay pronounced for onset and offset of active phase (ANOVA, p < 0.05), while the adults had the active phase more adjusted to light phase. Under artificial LD, there was phase advance and greater adjustment of onset and offset of activity in relation to the LD in the aged (ANOVA, p < 0.05). In LL, there was a positive correlation between age and the endogenous period () in the first 20 days (Spearman correlation, p < 0.05), with prolonged  held in two aged animals. In this condition, most adults showed free-running period of the circadian activity rhythm with  < 24 h for the first 30 days and later on relative coordination mediated by auditory cues. In study 2, the cross-correlation analysis between the activity profiles of the animals in LL with control animals kept under natural LD, found that there was less social synchronization in the aged. With the resubmission to the LD, the resynchronization rate was slower in the aged (t-test; p < 0.05) and in just one aged animal there was a loss of resynchronization capability. According to the data set, it is suggested that the aging in marmosets may be related to: 1) lower amplitude and greater fragmentation of the activity, accompanied to phase delay with extension of period, caused by changes in a photic input, in the generation and behavioral expression of the CAR; 2) lower capacity of the circadian activity rhythm to photic synchronization, that can become more robust in artificial lighting conditions, possibly due to the higher light intensities at the beginning of the active phase due to the abrupt transitions between the light and dark phases; and 3) smaller capacity of non-photic synchronization for auditory cues from conspecifics, possibly due to reducing sensory inputs and responsiveness of the circadian oscillators to auditory cues, what can make the aged marmoset most vulnerable, as these social cues may act as an important supporting factor for the photic synchronization.

Radionuclides measured on 21 water bottle profiles during POLARSTERN cruise ANT-XXIV/3

Drake Passage is a major route for many water masses from the strong Antarctic Circumpolar Current. During the ANTXXIV-3 expedition (in 2008) the vertical distributions of dissolved and size-fractionated particulate 231Pa and thorium isotopes (230Th, 232Th and 234Th) were investigated in order to better define the scavenging regimes and the effects of the oceanic circulation on the fate of particulate material and on the Pa-Th distributions in the water column. The reversible scavenging-model applied to both 230Th and 234Th, in the upper 1500 m depth, gives estimates of the particle dynamics (settling velocities S~ 500-1300 m/y, adsorption and desorption rate constants of 0.1-0.4 1/y and 1-6 1/y respectively). Particulate 234Th/230Th activity ratio shows a depth dependence, with decreasing ratio with increasing depth in agreement with previous studies, but no relationship with particle size was found. 231Pa and thorium isotope fractionation and partition coefficients were investigated with particle size vs depth and latitude and appear to vary horizontally following a North-South gradient. This suggests that both radionuclides are mostly bound to the fine suspended particles. At Drake Passage, the 230Thxs distribution is controlled by a southward upwelling of deep water (clearly visible on the vertical section of total 230Thxs, defined as dissolved + particulate concentrations) and reversible-scavenging processes (linear increase of 230Thxs with increasing depth) with North of the Southern ACC Front, higher settling velocities and less adsorption/desorption cycles, than South of it. Distributions of dissolved and total 231Paxs also reflect the influence of the North-South upwelling but somehow this effect appears to be limited to the upper 1500 m depth of the water column. Below this depth, 231Paxs vertical profiles exhibit contrasted concentrations, with some high dissolved activities in the deep water of the stations in the northern part of the ACC and not South of the ACC. These N-S differences in dissolved 231Paxs were attributed to the different origins and scavenging history of the deep Pacific waters flowing across Drake Passage. Here at North, radionuclides-rich deep water originates from the Central Pacific, while at South, deep water derives from the Southern Pacific in which the observed low radionuclides concentrations are attributed to high opal abundance. South of the Drake Passage, high dissolved and particulate activities of 230Th and 232Th confirmed the intrusion of 230Th-rich Weddell Sea Deep Water (WSDW) close to the Antarctic Peninsula.

Methods for Imputing Missing Values and Synthesizing Confidential Values for Continuous and Magnitude Data

Abstract

Continuous variable is one of the major data types collected by the survey organizations. It can be incomplete such that the data collectors need to fill in the missingness. Or, it can contain sensitive information which needs protection from re-identification. One of the approaches to protect continuous microdata is to sum them up according to different cells of features. In this thesis, I represents novel methods of multiple imputation (MI) that can be applied to impute missing values and synthesize confidential values for continuous and magnitude data.

The first method is for limiting the disclosure risk of the continuous microdata whose marginal sums are fixed. The motivation for developing such a method comes from the magnitude tables of non-negative integer values in economic surveys. I present approaches based on a mixture of Poisson distributions to describe the multivariate distribution so that the marginals of the synthetic data are guaranteed to sum to the original totals. At the same time, I present methods for assessing disclosure risks in releasing such synthetic magnitude microdata. The illustration on a survey of manufacturing establishments shows that the disclosure risks are low while the information loss is acceptable.

The second method is for releasing synthetic continuous micro data by a nonstandard MI method. Traditionally, MI fits a model on the confidential values and then generates multiple synthetic datasets from this model. Its disclosure risk tends to be high, especially when the original data contain extreme values. I present a nonstandard MI approach conditioned on the protective intervals. Its basic idea is to estimate the model parameters from these intervals rather than the confidential values. The encouraging results of simple simulation studies suggest the potential of this new approach in limiting the posterior disclosure risk.

The third method is for imputing missing values in continuous and categorical variables. It is extended from a hierarchically coupled mixture model with local dependence. However, the new method separates the variables into non-focused (e.g., almost-fully-observed) and focused (e.g., missing-a-lot) ones. The sub-model structure of focused variables is more complex than that of non-focused ones. At the same time, their cluster indicators are linked together by tensor factorization and the focused continuous variables depend locally on non-focused values. The model properties suggest that moving the strongly associated non-focused variables to the side of focused ones can help to improve estimation accuracy, which is examined by several simulation studies. And this method is applied to data from the American Community Survey.

Étude d’évaluabilité et évaluation des processus et des effets d’un programme de prévention de l’usage de substances psychoactives lors de la transition primaire-secondaire

La mise en œuvre d’activités de prévention de la consommation de substances psychoactives (SPA) (tabac, alcool et cannabis) en milieu scolaire est une stratégie couramment utilisée pour rejoindre un grand nombre de jeunes. Ces activités s’inspirent, soit de programmes existant, soit d’innovations dictées par le contexte d’implantation ou l’existence de données de recherche. Dans un cas comme dans l’autre, l’évaluation de ces programmes représente la meilleure voie pour mesurer leur efficacité et/ou connaître comment ceux-ci sont implantés. C’est cet impératif qui a motivé une commission scolaire du Québec a recommandé l’évaluation de l’Intervention en Réseau (IR), un programme développé en vue de retarder l’âge d’initiation et de réduire la consommation problématique de SPA chez les élèves. Ce programme adopte une approche novatrice avec pour principal animateur un intervenant pivot (IP) qui assure le suivi des élèves de la 5e année du primaire jusqu’en 3e secondaire. Inspiré des modèles en prévention de la santé et de l’Approche École en santé (AES), le rôle de l’IP ici se démarque de ceux-ci. Certes, il est l’interface entre les différents acteurs impliqués et les élèves mais dans le cadre du programme IR, l’IP est intégré dans les écoles primaires et secondaires qu’il dessert. C’est cet intervenant qui assure la mobilisation des autres acteurs pour la mise en œuvre des activités. Cette thèse vise à rendre compte de ce processus d’évaluation ainsi que des résultats obtenus. L’approche d’évaluation en est une de type participatif et collaboratif avec des données quantitatives et qualitatives recueillies par le biais de questionnaires, d’entrevues, de groupes de discussion, d’un journal de bord et de notes de réunions. Les données ont été analysées dans le cadre de trois articles dont le premier concerne l’étude d’évaluabilité (ÉÉ) du programme. Les participants de cette ÉÉ sont des acteurs-clés du programme (N=13) rencontrés en entrevues. Une analyse documentaire (rapports et journal de bord) a également été effectuée. Cette ÉÉ a permis de clarifier les intentions des initiateurs du programme et les objectifs poursuivis par ces derniers. Elle a également permis de rendre la théorie du programme plus explicite et de développer le modèle logique, deux éléments qui ont facilité les opérations d’évaluation qui ont suivi. Le deuxième article porte sur l’évaluation des processus en utilisant la théorie de l’acteur-réseau (TAR) à travers ses quatre moments du processus de traduction des innovations (la problématisation, l’intéressement, l’enrôlement et la mobilisation des alliés), l’analyse des controverses et du rôle des acteurs humains et non-humains. Après l’analyse des données obtenues par entrevues auprès de 19 informateurs-clés, les résultats montrent que les phases d’implantation du programme passent effectivement par les quatre moments de la TAR, que la gestion des controverses par la négociation et le soutien était nécessaire pour la mobilisation de certains acteurs humains. Cette évaluation des processus a également permis de mettre en évidence le rôle des acteurs non-humains dans le processus d’implantation du programme. Le dernier article concerne une évaluation combinée des effets (volet quantitatif) et des processus (volet qualitatif) du programme. Pour le volet quantitatif, un devis quasi-expérimental a été adopté et les données ont été colligées de façon longitudinale par questionnaires auprès de 901 élèves de 5e et 6e année du primaire et leurs enseignants de 2010 à 2014. L’analyse des données ont montré que le programme n’a pas eu d’effets sur l’accessibilité et les risques perçus, l’usage problématique d’alcool et la polyconsommation (alcool et cannabis) chez les participants. Par contre, les résultats suggèrent que le programme pourrait favoriser la réduction du niveau de consommation et retarder l’âge d’initiation à l’alcool et au cannabis. Ils suggèrent également un effet potentiellement positif du programme sur l’intoxication à l’alcool chez les élèves. Quant au volet qualitatif, il a été réalisé à l’aide d’entrevues avec les intervenants (N=17), de groupes de discussion avec des élèves du secondaire (N=10) et d’une analyse documentaire. Les résultats montrent que le programme bénéficie d’un préjugé favorable de la part des différents acteurs ayant participé à l’évaluation et est bien acceptée par ces derniers. Cependant, le roulement fréquent de personnel et le grand nombre d’écoles à suivre peuvent constituer des obstacles à la bonne marche du programme. En revanche, le leadership et le soutien des directions d’écoles, la collaboration des enseignants, les qualités de l’IP et la flexibilité de la mise en œuvre sont identifiés comme des éléments ayant contribué au succès du programme. Les résultats et leur implication pour les programmes et l’évaluation sont discutés. Enfin, un plan de transfert des connaissances issues de la recherche évaluative est proposé.

Modèles de dépendance hiérarchique pour l’évaluation des passifs et la tarification en actuariat

Dans cette thèse on s’intéresse à la modélisation de la dépendance entre les risques en assurance non-vie, plus particulièrement dans le cadre des méthodes de provisionnement et en tarification. On expose le contexte actuel et les enjeux liés à la modélisation de la dépendance et l’importance d’une telle approche avec l’avènement des nouvelles normes et exigences des organismes réglementaires quant à la solvabilité des compagnies d’assurances générales. Récemment, Shi et Frees (2011) suggère d’incorporer la dépendance entre deux lignes d’affaires à travers une copule bivariée qui capture la dépendance entre deux cellules équivalentes de deux triangles de développement. Nous proposons deux approches différentes pour généraliser ce modèle. La première est basée sur les copules archimédiennes hiérarchiques, et la deuxième sur les effets aléatoires et la famille de distributions bivariées Sarmanov. Nous nous intéressons dans un premier temps, au Chapitre 2, à un modèle utilisant la classe des copules archimédiennes hiérarchiques, plus précisément la famille des copules partiellement imbriquées, afin d’inclure la dépendance à l’intérieur et entre deux lignes d’affaires à travers les effets calendaires. Par la suite, on considère un modèle alternatif, issu d’une autre classe de la famille des copules archimédiennes hiérarchiques, celle des copules totalement imbriquées, afin de modéliser la dépendance entre plus de deux lignes d’affaires. Une approche avec agrégation des risques basée sur un modèle formé d’une arborescence de copules bivariées y est également explorée. Une particularité importante de l’approche décrite au Chapitre 3 est que l’inférence au niveau de la dépendance se fait à travers les rangs des résidus, afin de pallier un éventuel risque de mauvaise spécification des lois marginales et de la copule régissant la dépendance. Comme deuxième approche, on s’intéresse également à la modélisation de la dépendance à travers des effets aléatoires. Pour ce faire, on considère la famille de distributions bivariées Sarmanov qui permet une modélisation flexible à l’intérieur et entre les lignes d’affaires, à travers les effets d’années de calendrier, années d’accident et périodes de développement. Des expressions fermées de la distribution jointe, ainsi qu’une illustration empirique avec des triangles de développement sont présentées au Chapitre 4. Aussi, nous proposons un modèle avec effets aléatoires dynamiques, où l’on donne plus de poids aux années les plus récentes, et utilisons l’information de la ligne corrélée afin d’effectuer une meilleure prédiction du risque. Cette dernière approche sera étudiée au Chapitre 5, à travers une application numérique sur les nombres de réclamations, illustrant l’utilité d’un tel modèle dans le cadre de la tarification. On conclut cette thèse par un rappel sur les contributions scientifiques de cette thèse, tout en proposant des angles d’ouvertures et des possibilités d’extension de ces travaux.