Marginal Posterior Probability Research Articles

Abstract 6145: A trans-ethnic meta-analysis identified additional novel genomic loci associated with cervical cancer

Abstract Introduction: Cervical cancer is the second most common cause of cancer-related deaths among women worldwide. Genome-wide association studies (GWAS) have identified genetic variants associated with cervical cancer. However, these discoveries have been limited by the small sample sizes of the individual cohorts. To identify additional novel loci associated with cervical cancer, we performed a trans-ethnic meta-analysis. Methods: GWAS summary data for trans-ethic meta-analysis were identified in women of European ancestry from the UK Biobank, East Asian ancestry from Biobank Japan, and African ancestries from the pan-UK Biobank and Johannesburg cancer study. Meta-analyses were performed using fixed effect inverse variance weighted method implemented in GWAMA. To localize putative genomic loci associated with cervical cancer, we performed fine mapping using the Bayesian approach by calculating the marginal posterior probability of causality for each single nucleotide polymorphisms (SNPs) and 99% credible set size. Results: Our meta-analysis identified four additional novel genomic loci associated with cervical cancer. These loci include: rs147778467 (odds ratio [OR] = 1.12, 95% confidence interval [CI] = 1.08-1.16, P-value = 3.12 x 10-08) in RGS21, rs53150478 (OR =1.03, 95%CI = 1.01-1.05, P-value = 4.48 x 10-08) in MAN1C1, rs113842373 (OR = 0.68, 95%CI = 0.56-0.81, P-value = 2.31 x 10-09) in LINCO1850 and rs75372437 (OR = 0.80, 95%CI = 0.72-0.88, P-value = 3.32 x 10-08) in RNU6-356P. These loci are implicated in carcinogenic pathways. Moreover, our Bayesian fine mapping identified five loci with a marginal posterior probability of causality > 1 and reduced the 99% credible set sizes. Conclusion: We identified additional novel loci associated with cervical cancer and our fine mapping identified loci with a high posterior probability of being causal. Citation Format: Abram Bunya Kamiza, Jean-Tristan Brandenburg, Michele Ramsay, Christopher Mathew. A trans-ethnic meta-analysis identified additional novel genomic loci associated with cervical cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6145.

A Gaussian mixture multiple-model belief propagation filter for multisensor-multitarget tracking

This paper presents a novel Gaussian mixture multi-model belief propagation (GMM-BP) filter for maneuvering multitarget tracking with multiple sensors. The filter is built upon the BP-based multisensor-multitarget tracking scheme, enabling accurate estimation of target numbers and states. It assumes linear Gaussian target motion, birth process, and sensor measurement models and utilizes the Gaussian mixture model to represent the target's marginal posterior probability density function. Additionally, multiple models are incorporated for target motion, enhancing the maneuvering target tracking capability. Closed-form recursions for means, covariances, and weights of Gaussian components in the filter are derived. To enhance scalability with the number of sensors, a simplified multi-sensor data fusion process is proposed, thereby preserving a linear relationship between Gaussian components and the number of sensors. The performance of the filter was evaluated in a three-dimensional (3D) simulated environment. Compared to the particle-implemented BP filter and the labeled multi-Bernoulli (LMB) filter, the proposed GMM-BP filter exhibited a significant improvement in computational efficiency while maintaining a high level of tracking accuracy.

Joint trans-dimensional inversion for water-column sound speed and seabed geoacoustic models.

This letter considers joint estimation of the water-column sound-speed profile (SSP) and seabed geoacoustic model through Bayesian inversion of ocean-acoustic data. The inversion is formulated in terms of separate trans-dimensional models for the water column (as an unknown number of nodes of a piecewise-continuous SSP) and seabed (as an unknown number of uniform layers) to intrinsically parameterize each according to the information content of the data. The inversion estimates marginal posterior probability profiles, quantifying the resolution of water-column and seabed structure. To validate the proposed method, modal-dispersion data from the New England Mud Patch, collected using hand-deployable systems, are considered.

Winners from Winners: A Tale of Risk Factors

Starting from twelve distinct factors from the recent literature, plus twelve principal components (PCs) of anomalies unexplained by the initial factors, a Bayesian comparison of approximately seventeen million models in terms of marginal likelihoods and posterior model probabilities shows that {Mkt, MOM, IA, ROE, MGMT, PERF, PEAD, FIN}, plus the nonconsecutive principal components, {[Formula: see text]} are the best supported risk factors. Pricing tests and annualized out-of-sample Sharpe ratios for tangency portfolios suggest that this asset pricing model should be used for computing expected returns, assessing investment strategies and building portfolios. This paper was accepted by Lukas Schmid, finance. Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2023.4668 .

The Bayesian Estimate of Vector Autoregressive Model Parameters Adopt Informative Prior Information

This research included the bayesian estimate for vector Autoregressive model with rank (p) in addition to statistical tests and predict Bayesian when the random error of model followed generalized multivariate modified Bessel distribution. The prior information about the parameters of model is represented by probability distributions belong to conjugate families. It found that the posterior marginal probability distribution for parameters matrix (Φ) is a Matrix t distribution, The posterior marginal probability distribution of covariance matrix (Σ) is uncommon and the predictive probability distribution of future observations vector is multivariate t distribution.

The effect of interviewer-respondent age difference on the reporting of sexual activity in the Demographic and Health Surveys: Analysis of data from 21 countries.

Interviewer effects can have consequential impacts on survey data, particularly for reporting sensitive attitudes and behaviours such as sexual activity and drug use, yet these effects remain understudied in low- and middle-income countries. The Demographic and Health Surveys (DHS) present a unique opportunity to study interviewer effects on the self-report of sensitive topics in low- and middle-income countries by including interviewer characteristics data. This paper aims to narrow the gap in research on interviewer effects by studying the effects that age difference between interviewer and respondent and interviewer survey experience have on the reporting of ever having sexual intercourse. We used DHS data from 91 066 women and 56 336 men in 21 countries where the standard DHS was implemented among all women of reproductive age, and interviewer characteristics were included in the data set. Using a Bayesian cross-classified model with random intercepts for interviewer and cluster, we assessed whether the effect of an age difference of 10 years or greater was associated with a difference in self-report of ever having sexual intercourse, adjusting for respondent demographics. There was a meaningful association between an age difference of greater than ten years and reporting of ever having had sexual intercourse in most countries for both genders after adjusting for interviewer age and experience, rural or urban cluster, and individual-level characteristics. Among women, the marginal posterior probability of reporting ever having sexual intercourse if the interviewer was ten years or more years older was lower for 17 of 19 countries (countries ranged from -12.50 to 3.90 percentage points). Among men, the marginal posterior probability was lower for 16 of 20 countries, ranging from -18.30 to 17.10 percentage points. In most countries, women and men were less likely to report ever having sexual activity if the interviewer was ten or more years older than them, adjusting for potential confounders. These findings have important implications for interpreting numerous sexual health indicators, such as unmet family planning needs and human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) risk. Survey administrators may consider more careful interviewer-respondent characteristic matching or novel approaches like Audio Computer Assisted Self Interview to minimize interviewer-induced variance.

A subsampling approach for Bayesian model selection

It is common practice to use Laplace approximations to decrease the computational burden when computing the marginal likelihoods in Bayesian versions of generalised linear models (GLM). Marginal likelihoods combined with model priors are then used in different search algorithms to compute the posterior marginal probabilities of models and individual covariates. This allows performing Bayesian model selection and model averaging. For large sample sizes, even the Laplace approximation becomes computationally challenging because the optimisation routine involved needs to evaluate the likelihood on the full dataset in multiple iterations. As a consequence, the algorithm is not scalable for large datasets. To address this problem, we suggest using stochastic optimisation approaches, which only use a subsample of the data for each iteration. We combine stochastic optimisation with Markov chain Monte Carlo (MCMC) based methods for Bayesian model selection and provide some theoretical results on the convergence of the estimates for the resulting time-inhomogeneous MCMC. Finally, we report results from experiments illustrating the performance of the proposed algorithm.

Loopy sum-product algorithm based joint detection, tracking and classification of extended objects with analytic implementations

This paper addresses the problem of joint detection, tracking and classification (JDTC) of multiple extended objects (EOs), where each EO is restricted to be an ellipse modeled as a random matrix, and the classification of an EO is obtained on the basis of its size information. In order to reduce the computational complexity of data association and calculating the marginal posterior probability density function (PDF), the loopy sum-product algorithm (LSPA) is carried out based on a suitably devised factor graph. The goal of LSPA-based JDTC is achieved by iteratively computing the belief of each EO which is used to approximate the marginal posterior. Furthermore, in order to get a closed-form implementation, the belief of each EO is modeled via a sum of belief of each class, which is further modeled as the class probability multiplied by a mixture Gamma-Gaussian-inverse-Wishart (GGIW) distribution conditional on the class. The details of implementation are discussed in this paper and, finally, the effectiveness of the proposed approach is assessed via simulation experiments.

Direct Sequential Simulation for spherical linear inverse problems

We present a method for obtaining efficient probabilistic solutions to geostatistical and linear inverse problems in spherical geometry. Our Spherical Direct Sequential Simulation (SDSSIM) framework combines information from possibly noisy observations, that provide either point information on the model or are related to the model by a linear averaging kernel, and statistics derived from a-priori information. It generates realizations from marginal posterior probability distributions of model parameters that are not limited to be Gaussian. We avoid the restriction to Cartesian geometry built into many existing geostatistical simulation codes, and work instead with grids in spherical geometry relevant to problems in Earth and Space sciences.We demonstrate our scheme using a synthetic example, showing that it produces realistic posterior realizations consistent with the known solution while fitting observations within their uncertainty and reproducing the distribution of model parameters and covariance statistics of a-priori models. Secondly, we present an application to real satellite observations, estimating the posterior probability distribution for the geomagnetic field at the core–mantle boundary. Our results reproduce well-known features of the core–mantle boundary magnetic field, and also allow probabilistic investigations of the magnetic field morphology. Small-length scale features in the posterior realizations are not determined by the observations but match the covariance statistics extracted from geodynamo simulations. The framework presented here represents a step towards more general approaches to probabilistic inversion in spherical geometry.

Bayesian Inference of a Non normal Multivariate Partial Linear Regression Model

This research includes the Bayesian estimation of the parameters of the multivariate partial linear regression model when the random error follows the matrix-variate generalized modified Bessel distribution and found the statistical test of the model represented by finding the Bayes factor criterion, the predictive distribution under assumption that the shape parameters are known. The prior distribution about the model parameters is represented by non-informative information, as well as the simulate on the generated data from the model by a suggested way based on different values ​​of the shape parameters, the kernel function used in the generation was a Gaussian kernel function, the bandwidth (Smoothing) parameter was according to the rule of thumb. It found that the posterior marginal probability distribution of the location matrix and the predictive probability distribution is a matrix-t distribution with different parameters, the posterior marginal probability distribution of the scale matrix is proper distribution but it does not belong to the conjugate family, Through the Bayes factor criterion, it was found that the sample that was used in the generation process was drawn from a population that does not belong to the generalized modified Bessel population.

Bayesian maximum entropy for geoacoustic inversion using ship-radiated noise

Several acoustic signals from merchant ships, also known as ships of opportunity (SOOs), are employed for geacoustic inversion of the seabed on the New England Mud Patch Area using a Bayesian maximum entropy (BME) approach. The SOOs were recorded by three vertical line arrays deployed at different locations in the Seabed Characterization Experiment in 2017 (SBCEX2017). An effective seabed composed of a mud layer on top of a sandy layer is used for this work. For each SOO, the space of possible hypotheses H is uniformly sampled using Monte Carlo. For each hypothesis in H containing a set of geoacoustic parameters, a SOO spectrogram is simulated using the range independent normal mode model ORCA where the squared error between the measured and simulated SOO is calculated. BME is then applied to compute the marginal conditional posterior probability distribution of the geoacoustic parameters where the probability density function (PDF) provides the current state of knowledge. The inferred PDFs are consistent across all the measured SOOs used for this research and agree with previous geoacoustic inversions in the SBCEX2017 [Work supported by ONR Contract No. N00014-19-C-2001.]

Joint estimation of water-column and seabed properties by inversion of modal-dispersion data

This paper considers the joint estimation of water-column sound-speed profiles (SSP) and seabed geoacoustic models through Bayesian inversion of modal-dispersion data (arrival time as a function of frequency from warping time-frequency analysis). The inversion is formulated in terms of separate trans-dimensional models for the water column and seabed to intrinsically parametrize each according to the information content of the data. An efficient reversible-jump Markov-chain Monte Carlo algorithm is applied to estimate marginal posterior probability profiles, quantifying the resolution of SSP and seabed structure. Measured and simulated dispersion data, as well as varying levels of prior information, are considered to examine the ability to estimate water-column and seabed profiles, and to investigate relationships between these.

Bayesian Geophysical Inversion Using Invertible Neural Networks

AbstractConstraining geophysical models with observed data usually involves solving nonlinear and nonunique inverse problems. Neural mixture density networks (MDNs) provide an efficient way to estimate Bayesian posterior marginal probability density functions (pdf's) that represent the nonunique solution. However, it is difficult to infer correlations between parameters using MDNs, and in turn to draw samples from the posterior pdf. We introduce an alternative to resolve these issues: invertible neural networks (INNs). These are simultaneously trained to represent uncertain forward functions and to solve Bayesian inverse problems. In its usual form, the method does not account for uncertainty caused by data noise and becomes less effective in high dimensionality. To overcome these issues, in this study, we include data uncertainties as additional model parameters, and train the network by maximizing the likelihood of the data used for training. We apply the method to two types of imaging problems: One‐dimensional surface wave dispersion inversion and two‐dimensional travel time tomography, and we compare the results to those obtained using Monte Carlo and MDNs. Results show that INNs provide comparable posterior pdfs to those obtained using Monte Carlo, including correlations between parameters, and provide more accurate marginal distributions than MDNs. After training, INNs estimate posterior pdfs in seconds on a typical desktop computer. Hence they can be used to provide efficient solutions for repeated inverse problems using different data sets. Also even accounting for training time, our results show that INNs can be more efficient than Monte Carlo methods for solving inverse problems only once.

Archeomagnetic intensity investigations of French medieval ceramic workshops: Contribution to regional field modeling and archeointensity-based dating

Seven new archeointensity data are obtained through the analysis of groups of pottery and kiln fragments from ceramic workshops unearthed in France, precisely dated from the High Middle Ages. The measurements are carried out using the Triaxe magnetometer, following a dedicated experimental protocol that takes into account the effects of anisotropy and cooling rate (CR) on thermoremanent magnetization acquisition. The new data are consistent with the evolution of intensity variations described by our previous data obtained in France and Northern Italy, which display between the 5th and 10th c. a pronounced camel-back shape. In particular, they provide supporting evidence of an intensity minimum that occurred around the transition between the 7th and 8th century. These data, combined with a selection of previously published results within a 700 km radius of Beaune and re-examined based on CR correction, formed the basis of new regional mean intensity variation curves based on two independent modeling approaches. The first algorithm developed by Thébault and Gallet (2010) based on bootstrapping and now irregularly spaced knots according to the data distribution gives rather smooth intensity variations, while the second approach proposed by Livermore et al. (2018) based on a transdimensional Bayesian technique shows more abrupt variations with sometimes stronger amplitudes. We explore the dating potential of these two variations curves, which have an unprecedented resolution, by studying two medieval pottery workshops. Six fragment groups (three per workshop) are analyzed using the Triaxe protocol, providing mean archeointensity values for each of the two sites. Two different procedures are used for their dating, either by comparing the intensity value to be dated with the reference intensity variation curves obtained from the two modeling techniques or by analyzing the marginal posterior probability distribution of the age values derived from the method of Livermore et al. (2018). For France, the two techniques yield very similar results. The archeointensity dating results combined with archeological arguments and radiocarbon data, make it possible to better constrain the age of the end of activity of the two workshops. Archeointensity investigation of displaced materials thus appears as an effective means to obtain original chronological constraints on the age of their production, paving the way for a wide range of complementary research on Medieval pottery.

Consistent Sparse Deep Learning: Theory and Computation

Deep learning has been the engine powering many successes of data science. However, the deep neural network (DNN), as the basic model of deep learning, is often excessively over-parameterized, causing many difficulties in training, prediction and interpretation. We propose a frequentist-like method for learning sparse DNNs and justify its consistency under the Bayesian framework: the proposed method could learn a sparse DNN with at most connections and nice theoretical guarantees such as posterior consistency, variable selection consistency and asymptotically optimal generalization bounds. In particular, we establish posterior consistency for the sparse DNN with a mixture Gaussian prior, show that the structure of the sparse DNN can be consistently determined using a Laplace approximation-based marginal posterior inclusion probability approach, and use Bayesian evidence to elicit sparse DNNs learned by an optimization method such as stochastic gradient descent in multiple runs with different initializations. The proposed method is computationally more efficient than standard Bayesian methods for large-scale sparse DNNs. The numerical results indicate that the proposed method can perform very well for large-scale network compression and high-dimensional nonlinear variable selection, both advancing interpretable machine learning.

Stochastic model updating of rotor support parameters using Bayesian approach

Uncertainties due to assembling, installation, and operational conditions are extensively involved in rotating systems' parameters. Stochastic characteristics of these parameters may seriously affect rotating systems' vibrational characteristics, such as their critical speeds and vibration amplitudes. These effects make the variability of parameters in the modeling of rotary machines inevitable. In rotating machinery, material properties and geometric parameters of the rotor, bearing characteristics and supports stiffness determine the system's dynamic response. Stochastic model updating methods consider model response variability and allocate them to the model parameters; however, they are not commonly employed in rotor dynamics, and deterministic approaches are still prevalent in this field. Due to the cost and efforts needed to set up experiments and obtain outcomes that reflect the machine's actual characteristics, stochastic updating practices of industrial rotating systems are rarely reported in the literature. This paper adopts an appropriate parameter selection procedure and suitable sampling strategy for stochastic model updating to investigate variability in the dynamic behavior of a complex turbo compressor rotor-bearing-support system, leading to successful parameter identification results. The compressor rotor is mounted on hydrodynamic journal bearings with speed-dependent stiffness and damping. Due to the rotating system complex model, a variance-based global sensitivity method is employed for parameter selection to eliminate non-influential parameters in the model updating and to alleviate updating complexity and computational burden. The Bayesian approach in the stochastic model updating is applied to estimate parameter uncertainty in the rotor with speed-dependent characteristics. Advanced Markov chain Monte Carlo sampling method using delayed rejection adaptive Metropolis algorithm is employed in the stochastic model updating. The updating procedure obtains marginal posterior probabilities of parameters, and uncertain parameter distributions are evaluated using the maximum entropy criterion.

Assessing poison ivy (Toxicodendron radicans) presence and functional traits in relation to land cover and biophysical factors

ABSTRACT Understanding species distributions remains central to research in ecology and biogeography. Emphasis is placed on the spatial presence/absence of plants as related to underlying environmental factors; however, distributions result from both abiotic factors and adaptations to the abiotic environment. We analyzed poison ivy (Toxicodendron radicans (L.) Kuntze), which has high intraspecies variability in the plant functional trait of growth habit, to assess implications of plant strategies for spatial distributions. Our objectives were to: 1) determine whether anthropogenic habitats are statistically overrepresented in poison ivy incidence and growth habits (shrub, climbing liana, crawling liana), and 2) model biophysical parameters that constitute preferred poison ivy habitat and the three growth habits. We collected poison ivy field data along a trail-transect with corresponding geospatially indexed parameters, subsequently analyzed using Bayesian spatial modeling. Model results revealed poison ivy preference for human-modified habitat; Developed and Planted/Cultivated land use categories showed the largest marginal posterior probabilities for crawling and climbing lianas. Increasing temperature and elevation preferentially benefit the climbing growth habit. Our results suggest that variability in functional traits may impact species geographic distributions by expanding niche breadth. Incorporation of functional traits may thus advance predictive niche models of species distributions.

On the Bayesian network based data mining framework for the choice of appropriate time scale for regional analysis of drought Hazard

Data mining has a significant role in hyrdrologic research. Among several methods of data mining, Bayesian network theory has great importance and wide applications as well. The drought indices are very useful tools for drought monitoring and forecasting. However, the multi-scaling nature of standardized type drought indices creates several problems in data analysis and reanalysis at regional level. This paper presents a novel framework of data mining for hydrological research—the Bayesian Integrated Regional Drought Time Scale (BIRDts). The mechanism of BIRDts gives effective and sufficient time scales by considering dependency/interdependency probabilities from Bayesian network algorithm. The resultant time scales are proposed for further investigation and research related to the hydrological process. Application of the proposed method consists of 46 meteorological stations of Pakistan. In this research, we have employed Standardized Precipitation Temperature Index (SPTI) drought index for 1-, 3-, 6-, 9-, 12-, 24-, and ()month time scales. Outcomes associated with this research show that the proposed method has rationale to aggregate time scales at regional level by configuring marginal posterior probability as weights in the selection process of effective drought time scales.

Probability intervals of toxicity and efficacy design for dose-finding clinical trials in oncology.

Immunotherapy, gene therapy or adoptive cell therapies, such as the chimeric antigen receptor+ T-cell therapies, have demonstrated promising therapeutic effects in oncology patients. We consider statistical designs for dose-finding adoptive cell therapy trials, in which the monotonic dose-response relationship assumed in traditional oncology trials may not hold. Building upon a previous design called "TEPI", we propose a new dose finding method - Probability Intervals of Toxicity and Efficacy (PRINTE), which utilizes toxicity and efficacy jointly in making dosing decisions, does not require a pre-elicited decision table and at the same time can handle Ockham's razor properly in the statistical inference. We show that optimizing the joint posterior expected utility of toxicity and efficacy under a 0-1 loss is equivalent to maximizing the marginal model posterior probability in the two-dimensional probability space. An extensive simulation study under various scenarios are conducted and results show that PRINTE outperforms existing designs in the literature since it assigns more patients to optimal doses and less to toxic ones, and selects optimal doses with higher percentages. The simple and transparent features together with good operating characteristics make PRINTE an improved design for dose-finding trials in oncology trials.

Time-domain Markov chain Monte Carlo–based Bayesian damage detection of ballasted tracks using nonlinear ballast stiffness model

This article reports the development of a methodology for detecting ballast damage under a sleeper based on measured sleeper vibration following the Bayesian statistical system identification framework. To ensure the methodology is applicable under large amplitude vibration of the sleeper (e.g. under trainload), the nonlinear stress–strain behavior of railway ballast is considered. This, on one hand, significantly reduces the problem of modeling error, but, on the other hand, increases the number of uncertain model parameters. The uncertainty associated with the identified model parameters of the rail–sleeper–ballast system may be very high. To overcome this difficulty, the Markov chain Monte Carlo–based Bayesian model updating is adopted in the proposed methodology for the approximation of the posterior probability density function of uncertain model parameters. Owing to the nonlinear behavior of the system, the model updating is performed in the time domain instead of the modal domain. The applicability of the proposed damage detection methodology was first verified numerically using simulated impact hammer test data in two damaged cases perturbed with Gaussian white noise. Second, impact hammer tests of in situ sleepers in the full-scale in-door ballasted track test panel were carried out to collect data for the experimental verification of the proposed methodology. Artificial ballast damage was simulated under the target concrete sleeper by replacing normal-sized ballast particles (∼60 mm) by small-sized ballast particles (∼15 mm). The proposed methodology successfully identified the location and severity of ballast damage under the sleeper. From the calculated posterior marginal probability density functions of model parameters, one can quantify the uncertainties associated with the damage detection results. The proposed methodology is an essential step in the development of a long-term railway track health monitoring system utilizing train-induced vibration.