Abstract. Bayesian Model Averaging (BMA) is an application of Bayesian inference to the problems of model selection, combined estimation and prediction that produces a straightforward model choice criteria and less risky predictions. However, the application of BMA is not always straightforward, leading to diverse assumptions and situational choices on its different aspects. Despite the widespread application of BMA in the literature, there were not many accounts of these differences and trends besides a few landmark revisions in the late 1990s and early 2000s, therefore not taking into account any advancements made in the last 15 years. In this work, we present an account of these developments through a careful content analysis of 587 articles in BMA published between 1996 and 2014. We also develop a conceptual classification scheme to better describe this vast literature, understand its trends and future directions and provide guidance for the researcher interested in both the application and development of the methodology. The results of the classification scheme and content review are then used to discuss the present and future of the BMA literature.
In this article, we propose a class of zero-modified Poisson mixture models as an alternative to model overdispersed count data exhibiting inflation or deflation of zeros. A relevant feature of this class is that the zero modification can be incorporated using a zero truncation process and consequently, the proposed models can be expressed in the hurdle version. This procedure leads to the fact that the proposed models can be fitted without any previous information about the zero modification present in agiven dataset. A fully Bayesian approach has been considered for estimation and inference concerns. Three different simulation studies have been conducted to illustrate the performance of the developed methodology. The usefulness of the proposed class of models has been assessed by using three real datasets provided by the literature. A general model comparison with some well-known discrete distributions has been presented.
Biological and medical sciences are increasingly acknowledging the significance of gene co-expression-networks for investigating complex-systems, phenotypes or diseases. Typically, complex phenotypes are investigated under varying conditions. While approaches for comparing nodes and links in two networks exist, almost no methods for the comparison of multiple networks are available and—to best of our knowledge—no comparative method allows for whole transcriptomic network analysis. However, it is the aim of many studies to compare networks of different conditions, for example, tissues, diseases, treatments, time points, or species. Here we present a method for the systematic comparison of an unlimited number of networks, with unlimited number of transcripts: Co -expression Di fferential N etwork A nalysis (CoDiNA). In particular, CoDiNA detects links and nodes that are common, specific or different among the networks. We developed a statistical framework to normalize between these different categories of common or changed network links and nodes, resulting in a comprehensive network analysis method, more sophisticated than simply comparing the presence or absence of network nodes. Applying CoDiNA to a neurogenesis study we identified candidate genes involved in neuronal differentiation. We experimentally validated one candidate, demonstrating that its overexpression resulted in a significant disturbance in the underlying gene regulatory network of neurogenesis. Using clinical studies, we compared whole transcriptome co-expression networks from individuals with or without HIV and active tuberculosis (TB) and detected signature genes specific to HIV. Furthermore, analyzing multiple cancer transcription factor (TF) networks, we identified common and distinct features for particular cancer types. These CoDiNA applications demonstrate the successful detection of genes associated with specific phenotypes. Moreover, CoDiNA can also be used for comparing other types of undirected networks, for example, metabolic, protein-protein interaction, ecological and psychometric networks. CoDiNA is publicly available as an package in CRAN ( https://CRAN.R-project.org/package=CoDiNA ).
The primary goal of this paper is to introduce the zero-modified Poisson-Lindley regression model as an alternative to model overdispersed count data exhibiting inflation or deflation of zeros in the presence of covariates. The zero-modification is incorporated by considering that a zerotruncated process produces positive observations and consequently, the proposed model can be fitted without any previous information about the zeromodification present in a given dataset. A fully Bayesian approach based on the g-prior method has been considered for inference concerns. An intensive Monte Carlo simulation study has been conducted to evaluate the performance of the developed methodology and the maximum likelihood estimators. The proposed model was considered for the analysis of a real dataset on the number of bids received by 126 U.S. firms between 1978-1985, and the impact of choosing different prior distributions for the regression coefficients has been studied. A sensitivity analysis to detect influential points has been performed based on the Kullback-Leibler divergence. A general comparison with some well-known regression models for discrete data has been presented. IntroductionThe standard Poisson (P) distribution is the most adopted discrete model for the analysis of count data in several research fields, mainly due to its great simplicity and by having its computational implementation available for many statistical packages. However, it is well-known that such model is not a suitable choice for the analysis of counts in which the variance-to-mean ratio is not (at least) close to 1, that is, when the equidispersion property is violated. Apart from data transformation, the most popular way to circumvent such an issue is the use of finite mixture models (McLachlan and Peel, 2000) that can accommodate, for example, overdispersion (Karlis and Xekalaki, 2005). The Negative Binomial (N B) distribution (that may arise as a P mixture model by using a Gamma distribution for the continuous part) is undoubtedly the most famous alternative to model extra-P variability. However, the literature concerning discrete models, which can handle different
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.