Odds ratio

Résumé

L’odds ratio (OR), également appelé rapport des chances, rapport des cotes ou risque relatif rapproché, est une mesure statistique, souvent utilisée en épidémiologie, exprimant le degré de dépendance entre des variables aléatoires qualitatives. Il est utilisé en inférence bayésienne et en régression logistique, et permet de mesurer l'effet d'un facteur. Définition Lodds ratio se définit comme le rapport de la cote d'un événement arrivant à un groupe A d'individus, par exemple une maladie, avec celle du même événement arrivant à un groupe B d'individus. Les cotes sont à entendre comme celle d'un cheval de course dans un grand prix. Un cheval à 3 contre 1 a une chance sur 4 de gagner. L’odds ratio est proche du risque relatif lorsque le nombre d’événements est faible. En d'autres termes, si p est petit alors p/(1-p) est à peu près égal à p. Si p est la probabilité qu'un événement arrive dans le groupe A,

Source officielle

https://fr.wikipedia.org/wiki/Odds_ratio

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Publications associées (7)

This paper addresses the basic concepts of the packet switching in the optical domain and describes an analytical approach to evaluate the end-to-end performance of networks employing slotted (fixed length) optical packets. Thus, for a given topology and traffic matrix, the end-to-end cell loss ratio is analytically computed assuming an uncorrelated traffic. A network dimensioning method relying on this approach is also presented.

Springer2006

Chargement

Tumor control and radiobiological fingerprint after Gamma Knife radiosurgery for posterior fossa meningiomas: A series of 46 consecutive cases

Constantin Tuleasca

Introduction: Gamma Knife radiosurgery (GKR) can be a valuable treatment option for posterior cranial fossa meningiomas (PCFM). We retrospectively analyzed outcomes of GKR for PCFM.& nbsp;Methods: Were included forty-six patients with 47 PCFM. Primary endpoint was tumor control. Secondary endpoint was clinical improvement. Biologically effective dose (BED) was evaluated in relationship to primary and secondary outcomes. Mean marginal dose was 12.4 Gy (median 12, 12-14). Mean BED was 63.6 Gy (median 65, 49.1-88.3). Mean target volume (TV) was 2.21 cc (range 0.3-8.9 cc).& nbsp;Results: Overall tumor control rate was 93.6% (44/47) after mean follow-up of 47.8 months & PLUSMN; 28.46 months (median 45.5, range 6-108). Radiological progression-free survival at 5 years was 94%. Higher pretherapeutic TVs were predictive for higher likelihood of tumor progression (Odds ratio, OR 1.448, 95% confidence interval CI 1.001-2.093, p = 0.049). At last clinical follow-up, 28 patients (71.8%) remained stable, 10 (25.6%) improved and 1 patient (2.6%) worsened. Using logistic regression, the relationship between BED and clinical improvement was assessed (OR 0.903, standard error 0.59, coefficient 0.79-1.027, CI-0.10; 0.01; p = 0.14). The highest probability of clinical improvement corresponded to a range of BED values between 56 and 61 Gy.& nbsp;Conclusion: Primary GKR for PCFM is safe and effective. Higher pretherapeutic TV was predictor of volumetric progression. Highest probability of clinical improvement might correspond to a range of BED values between 56 and 61 Gy, although this was not statistically significant. The importance of BED should be further validated in larger cohorts, other anatomical locations and other pathologies.

2022

Chargement

, , ,

Surprise-based learning allows agents to rapidly adapt to nonstationary stochastic environments characterized by sudden changes. We show that exact Bayesian inference in a hierarchical model gives rise to a surprise-modulated trade-off between forgetting old observations and integrating them with the new ones. The modulation depends on a probability ratio, which we call the Bayes Factor Surprise, that tests the prior belief against the current belief. We demonstrate that in several existing approximate algorithms, the Bayes Factor Surprise modulates the rate of adaptation to new observations. We derive three novel surprise-based algorithms, one in the family of particle filters, one in the family of variational learning, and one in the family of message passing, that have constant scaling in observation sequence length and particularly simple update dynamics for any distribution in the exponential family. Empirical results show that these surprise-based algorithms estimate parameters better than alternative approximate approaches and reach levels of performance comparable to computationally more expensive algorithms. The Bayes Factor Surprise is related to but different from the Shannon Surprise. In two hypothetical experiments, we make testable predictions for physiological indicators that dissociate the Bayes Factor Surprise from the Shannon Surprise. The theoretical insight of casting various approaches as surprise-based learning, as well as the proposed online algorithms, may be applied to the analysis of animal and human behavior and to reinforcement learning in nonstationary environments.

MIT PRESS2021

Chargement

Afficher plus

Personnes associées

Aucun résultat

Unités associées

Chargement

Concepts associés (16)

En statistiques, la régression logistique ou modèle logit est un modèle de régression binomiale. Comme pour tous les modèles de régression binomiale, il s'agit d'expliquer au mieux une variable bin

L'épidémiologie est une discipline scientifique qui étudie les problèmes de santé dans les populations humaines, leur fréquence, leur distribution dans le temps et dans l’espace, ainsi que les fac

En probabilités et en statistique, la corrélation entre plusieurs variables aléatoires ou statistiques est une notion de liaison qui contredit leur indépendance. Cette corrélation est très souvent

Afficher plus

Concepts associés

Chargement

Cours associés (12)

This course is neither an introduction to the mathematics of statistics nor an introduction to a statistics program such as R. The aim of the course is to understand statistics from its experimental design and to avoid common pitfalls of statistical reasoning. There is space to discuss ongoing work.

This course covers topics in applied biostatistics, with an emphasis on practical aspects of data analysis using R statistical software. Topics include types of studies and their design and analysis, high dimensional data analysis (genetic/genomic) and other topics as time and interest permit.

This course teaches the basic techniques, methodologies, and practical skills required to draw meaningful insights from a variety of data, with the help of the most acclaimed software tools in the data science world (pandas, scikit-learn, Spark, etc.)

Afficher plus

Cours associés

Chargement

Séances de cours associées (21)

Séances de cours associées

Chargement