Bayesian Selection Research Articles

Massive Uncoordinated Multiple Access for Beyond 5G

Existing wireless communication systems have been mainly designed to provide substantial gain in terms of data rates. However, 5G and Beyond will depart from this scheme, with the objective not only to provide services with higher data rates. One of the main goals is to support massive machine-type communications (mMTC) in the Internet-of-Things (IoT) applications. Supporting massive uplink communications for devices with sporadic traffic pattern and short-packet size, as it is in many mMTC use cases, is a challenging task, particularly when the control signaling is not negligible in size compared to the payload. In addition, channel estimation becomes challenging for sporadic and short-packet transmission due to the limited number of employed pilots. In this paper, a new uplink multiple access (MA) scheme is proposed for mMTC, which can support a large number of uncoordinated IoT devices with short-packet and sporadic traffic. The proposed uplink MA scheme removes the overheads associated with the device identifier as well as pilots and preambles related to channel estimation. An alternative mechanism for device identification (DI) is employed, where a unique spreading code is dedicated to each IoT device as identifier. This unique code is simultaneously used for the spreading purpose and DI. Two IoT DI algorithms which employ sparse signal reconstruction methods are proposed to determine the active IoT devices prior to data detection. Specifically, the Bayesian information criterion model order selection method is employed to develop an IoT DI algorithm for unknown and time-varying activity rate. Our proposed MA scheme benefits from a new non-coherent nonlinear multiuser detection algorithm designed on the basis of unsupervised machine learning techniques to enable data detection without <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a priori</i> knowledge on channel state information. For performance improvement, an extension to multiple receive antennas through hard decision combining is proposed. The effectiveness of the proposed MA scheme for known and unknown activity rate and high overloading factor is supported by simulation results.

Early termination in single-parameter model phase II clinical trial designs using decreasingly informative priors.

To exchange the type of subjective Bayesian prior selection for assumptions more directly related to statistical decision making in clinician studies and trials, the decreasingly informative prior (DIP) is considered. We expand standard Bayesian early termination methods in one-parameter statistical models for Phase II clinical trials to include decreasingly informative priors (DIP). These priors are designed to reduce the chance of erroneously adapting trials too early by parameterize skepticism in an amount always equal to the unobserved sample size. We show how to parameterize these priors based on effective prior sample size and provide examples for common single-parameter models, include Bernoulli, Poisson, and Gaussian distributions. We use a simulation study to search through possible values of total sample sizes and termination thresholds to find the smallest total sample size (N) under admissible designs, which we define as having at least 80% power and no greater than 5% type I error rate. For Bernoulli, Poisson, and Gaussian distributions, the DIP approach requires fewer patients when admissible designs are achieved. In situations where type I error or power are not admissible, the DIP approach yields similar power and better-controlled type I error with comparable or fewer patients than other Bayesian priors by Thall and Simon. The DIP helps control type I error rates with comparable or fewer patients, especially for those instances when increased type I error rates arise from erroneous termination early in a trial.

Bayesian portfolio selection using VaR and CVaR

We study the optimal portfolio allocation problem from a Bayesian perspective using value at risk (VaR) and conditional value at risk (CVaR) as risk measures. By applying the posterior predictive distribution for the future portfolio return, we derive relevant quantities needed in the computations of VaR and CVaR, and express the optimal portfolio weights in terms of observed data only. This is in contrast to the conventional method where the optimal solution is based on unobserved quantities which are estimated. We also obtain the expressions for the weights of the global minimum VaR (GMVaR) and global minimum CVaR (GMCVaR) portfolios, and specify conditions for their existence. It is shown that these portfolios may not exist if the level used for the VaR or CVaR computation are too low. By using simulation and real market data, we compare the new Bayesian approach to the conventional plug-in method by studying the accuracy of the GMVaR portfolio and by analysing the estimated efficient frontiers. It is concluded that the Bayesian approach outperforms the conventional one, in particular at predicting the out-of-sample VaR.

Sequential Bayesian bandwidth selection for multivariate kernel regression with applications

A new Bayesian bandwidth selection procedure is proposed for nonparametric kernel estimates based on the sequential Monte Carlo method. Compared with the existing Bayesian bandwidth selector of Zhang et al. (2009), this new method can enhance the convergence to the global optimum with a substantially faster computation speed. In particular, the method offers an improved out-of-sample performance as shown by simulations. The bandwidth selector is applied to the option state price density, production function, and nonparametric relationship between oil and stock index returns; results indicate that our proposed method outperforms other methods in terms of the mean square error and log-likelihood in all applications.

A simulation study to examine the impact of recombination on phylogenomic inferences under the multispecies coalescent model.

Phylogenomic analyses under the multispecies coalescent model assume no recombination within locus and free recombination among loci. Yet, in real data sets intralocus recombination causes different sites of the same locus to have different genealogical histories so that the model is misspecified. The impact of recombination on various coalescent‐based phylogenomic analyses has not been systematically examined. Here, we conduct a computer simulation to examine the impact of recombination on several Bayesian analyses of multilocus sequence data, including species tree estimation, species delimitation (by Bayesian selection of delimitation models) and estimation of evolutionary parameters such as species divergence and introgression times, population sizes for modern and extinct species, and cross‐species introgression probabilities. We found that recombination, at rates comparable to estimates from the human being, has little impact on coalescent‐based species tree estimation, species delimitation and estimation of population parameters. At rates 10 times higher than the human rate, recombination may affect parameter estimation, causing positive biases in introgression times and ancestral population sizes, although species divergence times and cross‐species introgression probabilities are estimated with little bias. Overall, the simulation suggests that phylogenomic inferences under the multispecies coalescent model are robust to realistic amounts of intralocus recombination.

Imagination as a skill: A Bayesian proposal

In recent works, Kind (2020a, b) has argued that imagination is a skill, since it possesses the two hallmarks of skill: (i) improvability by practice, and (ii) control. I agree with Kind that (i) and (ii) are indeed hallmarks of skill, and I also endorse her claim that imagination is a skill in virtue of possessing these two features. However, in this paper, I argue that Kind’s case for imagination’s being a skill is unsatisfactory, since it lacks robust empirical evidence. Here, I will provide evidence for (i) by considering data from mental rotation experiments and for (ii) by considering data from developmental experiments. I conclude that imagination is a skill, but there is a further pressing question of how the cognitive architecture of imagination has to be structured to make this possible. I begin by considering how (ii) can be implemented sub-personally. I argue that this can be accounted for by positing a selection mechanism which selects content from memory representations to be recombined into imaginings, using Bayesian generation. I then show that such an account can also explain (i). On this basis, I hold that not only is imagination a skill, but that it is also plausibly implemented sub-personally by a Bayesian selection mechanism.

Hierarchical Bayesian Approach for Model Parameter Updating in Piezoelectric Energy Harvesters

This work proposes a hierarchical Bayesian framework to identify electromechanical properties of Piezoelectric Energy Harvesters (PEHs) and associated uncertainties based on experimental frequency response functions (FRFs). The framework allows the use of experimental data from multiple devices, potentially defined by different electromechanical properties. In the proposed hierarchical scheme, the FRF dispersion experimentally observed in groups of PEHs is explicitly modeled as a consequence of uncertainties in the model parameters rather than as a consequence of only the model prediction error typically used in classical Bayesian scheme. The Transitional Markov Chain Monte Carlo (TMCMC) method is used to establish the full posterior distribution of the model parameters. Preference towards selection of the hierarchical scheme is further confirmed by using Bayesian model class selection to compare the posterior probabilities of selecting the hierarchical or the classical scheme. The proposed framework is applied to identification of model parameters for both a single device and groups of devices. Results show that the proposed hierarchical scheme present significant advantages compared to other Bayesian based approaches for PEHs. First, it extracts more information about the model parameters using the same experimental observations (i.e., extract information at both device and group level); second, it accounts for the model parameter uncertainties across different devices; third, it could be used to identify objective priors for classical Bayesian approach.

Bayesian interaction selection model for multimodal neuroimaging data analysis.

Multimodality or multiconstruct data arise increasingly in functional neuroimaging studies to characterize brain activity under different cognitive states. Relying on those high-resolution imaging collections, it is of great interest to identify predictive imaging markers and intermodality interactions with respect to behavior outcomes. Currently, most of the existing variable selection models do not consider predictive effects from interactions, and the desired higher-order terms can only be included in the predictive mechanism following a two-step procedure, suffering from potential misspecification. In this paper, we propose a unified Bayesian prior model to simultaneously identify main effect features and intermodality interactions within the same inference platform in the presence of high-dimensional data. To accommodate the brain topological information and correlation between modalities, our prior is designed by compiling the intermediate selection status of sequential partitions in light of the data structure and brain anatomical architecture, so that we can improve posterior inference and enhance biological plausibility. Through extensive simulations, we show the superiority of our approach in main and interaction effects selection, and prediction under multimodality data. Applying the method to the Adolescent Brain Cognitive Development (ABCD) study, we characterize the brain functional underpinnings with respect to general cognitive ability under different memory loadconditions.

Belief correlations with parental vaccine hesitancy: Results from a national survey

We conducted a nationally representative survey of parents’ beliefs and self‐reported behaviors regarding childhood vaccinations. Using Bayesian selection among multivariate models, we found that beliefs, even those without any vaccine or health content, predicted vaccine‐hesitant behaviors better than demographics, social network effects, or scientific reasoning. The multivariate structure of beliefs combined many types of ideation that included concerns about both conspiracies and side effects. Although they are not strongly related to vaccine‐hesitant behavior, demographics were key predictors of beliefs. Our results support some of the previously proposed pro‐vaccination messaging strategies and suggest some new strategies not previously considered.

Objective Bayesian Edge Screening and Structure Selection for Ising Networks

The Ising model is one of the most widely analyzed graphical models in network psychometrics. However, popular approaches to parameter estimation and structure selection for the Ising model cannot naturally express uncertainty about the estimated parameters or selected structures. To address this issue, this paper offers an objective Bayesian approach to parameter estimation and structure selection for the Ising model. Our methods build on a continuous spike-and-slab approach. We show that our methods consistently select the correct structure and provide a new objective method to set the spike-and-slab hyperparameters. To circumvent the exploration of the complete structure space, which is too large in practical situations, we propose a novel approach that first screens for promising edges and then only explore the space instantiated by these edges. We apply our proposed methods to estimate the network of depression and alcohol use disorder symptoms from symptom scores of over 26,000 subjects.

National, regional, and global causes of mortality in 5–19-year-olds from 2000 to 2019: a systematic analysis

SummaryBackgroundInvestments in the survival of older children and adolescents (aged 5–19 years) bring triple dividends for now, their future, and the next generation. However, 1·5 million deaths occurred in this age group globally in 2019, nearly all from preventable causes. To better focus the attention of the global community on improving survival of children and adolescents and to guide effective policy and programmes, sound and timely cause of death data are crucial, but often scarce.MethodsIn this systematic analysis, we provide updated time-series for 2000–19 of national, regional, and global cause of death estimates for 5–19-year-olds with age-sex disaggregation. We estimated separately for countries with high versus low mortality, by data availability, and for four age-sex groups (5–9-year-olds [both sexes], 10–14-year-olds [both sexes], 15–19-year-old females, and 15–19-year-old males). Only studies reporting at least two causes of death were included in our analysis. We obtained empirical cause of death data through systematic review, known investigator tracing, and acquisition of known national and subnational cause of death studies. We adapted the Bayesian Least Absolute Shrinkage and Selection Operator approach to address data scarcity, enhance covariate selection, produce more robust estimates, offer increased flexibility, allow country random effects, propagate coherent uncertainty, and improve model stability. We harmonised all-cause mortality estimates with the UN Inter-agency Group for Child Mortality Estimation and systematically integrated single cause estimates as needed from WHO and UNAIDS.FindingsIn 2019, the global leading specific causes of death were road traffic injuries (115 843 [95% uncertainty interval 110 672–125 054] deaths; 7·8% [7·5–8·1]); neoplasms (95 401 [90 744–104 812]; 6·4% [6·1–6·8]); malaria (81 516 [72 150–94 477]; 5·5% [4·9–6·2]); drowning (77 460 [72 474–85 952]; 5·2% [4·9–5·5]); and diarrhoea (72 679 [66 599–82 002], 4·9% [4·5–5·3]). The leading causes varied substantially across regions. The contribution of communicable, maternal, perinatal, and nutritional conditions declined with age, whereas the number of deaths associated with injuries increased. The leading causes of death were diarrhoea (51 630 [47 206–56 235] deaths; 10·0% [9·5–10·5]) in 5–9-year-olds; malaria (31 587 [23 940–43 116]; 8·6% [6·6–10·4]) in 10–14-year-olds; self-harm (32 646 [29 530–36 416]; 13·4% [12·6–14·3]) in 15–19-year-old females; and road traffic injuries (48 757 [45 692–52 625]; 13·9% [13·3–14·3]) in 15–19-year-old males. Widespread declines in cause-specific mortality were estimated across age-sex groups and geographies in 2000–19, with few exceptions like collective violence.InterpretationChild and adolescent survival needs focused attention. To translate the vision into actions, more investments in the health information infrastructure for cause of death and in the related life-saving interventions are needed.FundingBill & Melinda Gates Foundation and WHO.

Bayesian subcohort selection for longitudinal covariate measurements in follow‐up studies

AbstractWe propose an approach for the planning of longitudinal covariate measurements in follow‐up studies where covariates are time‐varying. We assume that the entire cohort cannot be selected for longitudinal measurements due to financial limitations, and study how a subset of the cohort should be selected optimally, in order to obtain precise estimates of covariate effects in a survival model. In our approach, the study will be designed sequentially utilizing the data collected in previous measurements of the individuals as prior information. We propose using a Bayesian optimality criterion in the subcohort selections, which is compared with simple random sampling using simulated and real follow‐up data. Our work improves the computational approach compared to the previous research on the topic so that designs with several covariates and measurement points can be implemented. As an example we derive the optimal design for studying the effect of body mass index and smoking on all‐cause mortality in a Finnish longitudinal study. Our results support the conclusion that the precision of the estimates can be clearly improved by optimal design.

Which roads lead to depression in Latinas? A network analysis of prenatal depressive symptoms, discrimination, acculturative stress, and low birth weight.

Although immigrant mothers from some Latinx subgroups initially achieve healthy birth outcomes despite lower socioeconomic status, this advantage deteriorates across generations in the United States. Interpersonal discrimination and acculturative stress may interact with economic hardship to predict an intergenerational cascade of emotional and biological vulnerabilities, particularly perinatal depression. Network analyses may elucidate not only how and which psychosocial experiences relate to depressive symptoms, but which symptom-to-symptom relationships emerge. This study aims to understand (1) how economic, acculturative, and discrimination stressors relate to prenatal depression and low birth weight and (2) how Latinas may respond to and cope with stressors by exploring symptom-symptom and symptom-experience relationships. A sample of 151 pregnant Latinas (predominantly foreign-born and Mexican and Central American descent) completed the EPDS and psychosocial questionnaires (discrimination, acculturation, acculturative stress, economic hardship) during pregnancy (24-32 weeks). Birth weights were recorded from postpartum medical records. We created network models using the Extended Bayesian Information Criterion Graphical Least Absolute Shrinkage and Selection Operator to estimate the relationship between variables. Discrimination exposure connected psychosocial stressors to depressive symptoms, particularly worry, crying, sadness, and self-blame. Discrimination also revealed a connection between acculturation and low birth weight. Furthermore, younger age of migration and greater acculturation levels were correlated to greater discrimination stress and low birth weights. Perinatal research in Latinas must account not only for measures of cultural adaptation but recognize how developmental exposures across the life span, including discrimination, may be associated with adverse health trajectories for a mother and her child.

Correlated geometric models of order k and its application to intensive care unit and leprosy data.

Geometric models are used to analyse the discrete time until the occurrence of an event of interest (success or consecutive successes). In two real data sets, named leprosy and intensive care unit (ICU), the events correspond, respectively, to abandoning the clinical treatment of leprosy, where abandonment corresponds to four consecutive patient absences from treatment, and the patient's discharge from the ICU. The distribution proposed in this article, called the correlated geometric distribution of order k (or correlated k-order geometric distribution), , consists of including a correlation parameter in the geometric distribution of order k, thus considering the dependence between patient responses until the occurrence of the event. This model proves to be a better option for real data analysis where the effect of individual correlation is considered. The model is applied to real leprosy data to estimate the treatment abandonment probability. Bayesian methods are used to determine the parameter estimators of the models and to evaluate regression models. The covariates are related to the probability of the event by an appropriate link function chosen by Bayesian selection criteria. A diagnostic analysis evaluates the models fit by posterior randomized quantile residuals and influential observations by -divergence measures. This methodology is illustrated by simulation studies and real ICU admission data analysis. Studies show a good fit of the proposed model. Real data analyses also find that the probabilities of the event of interest can be overestimated or underestimated when modeled without considering the effect of dependency on the model.

Bayesian Sample Selection Model with Binary Travel Behavior for Addressing Residential Self-Selection

Bayesian Sample Selection Model with Binary Travel Behavior for Addressing Residential Self-Selection

Characterising a unique recreational hunting method: hound hunting of sambar deer (Cervus unicolor) in Victoria, Australia

Context Scent-trailing dogs (‘hounds’) are used to hunt introduced sambar deer (Cervus unicolor) in south-eastern Australia, but little is known about this recreational hunting technique. Aim The aim of this study was to characterise hound hunting of sambar deer in Victoria, Australia. Our study had three objectives as follows: (1) to report the hunting success of the technique; (2) to estimate pursuit times and distances; and (3) to investigate the landscape features associated with kill sites. Methods Data were collected from four hound-hunting crews during 2020–2021. GPS data from collars fitted to hounds were used to quantify hunt duration (min) and hunt distance (m). Logistic regression was used to estimate the effects of deer sex and hound pack size on the probability of hunting success. We used a Bayesian multinomial regression resource selection function (RSF) to identify the characteristics of sites where deer were killed (elevation, aspect, and distances to water and roads). Key results The four hunting teams ranged in size from 2 to 10 people and from one to eight hounds. Of 136 hunts, most (88%) harvested a deer, with pursued deer escaping on the other 12% of occasions. Pursuit times and distances were highly variable, with pursuit time &gt;60 min for 46% of hunts and pursuit distance &gt;5 km for 30% of hunts. The probability of killing a pursued male and female deer were similar, and there was not a positive relationship between pack size and hunt success. The RSF showed that both male and female deer were more likely to be killed on steeper slopes and closer to roads. Conclusions Successful hound hunting of sambar deer involves pursuits of considerable duration and requires a network of roads, with geographical features associated with deer kills sites being related to hunter access. Implications Hound hunting may be a useful wildlife management tool for land managers, but further studies are needed to assess its efficacy for achieving management goals.

Sparse linear mixed model selection via streamlined variational Bayes

Linear mixed models are a versatile statistical tool to study data by accounting for fixed effects and random effects from multiple sources of variability. In many situations, a large number of candidate fixed effects is available and it is of interest to select a parsimonious subset of those being effectively relevant for predicting the response variable. Variational approximations facilitate fast approximate Bayesian inference for the parameters of a variety of statistical models, including linear mixed models. However, for models having a high number of fixed or random effects, simple application of standard variational inference principles does not lead to fast approximate inference algorithms, due to the size of model design matrices and inefficient treatment of sparse matrix problems arising from the required approximating density parameters updates. We illustrate how recently developed streamlined variational inference procedures can be generalized to make fast and accurate inference for the parameters of linear mixed models with nested random effects and global-local priors for Bayesian fixed effects selection. Our variational inference algorithms achieve convergence to the same optima of their standard implementations, although with significantly lower computational effort, memory usage and time, especially for large numbers of random effects. Using simulated and real data examples, we assess the quality of automated procedures for fixed effects selection that are free from hyperparameters tuning and only rely upon variational posterior approximations. Moreover, we show high accuracy of variational approximations against model fitting via Markov Chain Monte Carlo sampling.

Developing a prediction model for segment joint opening in an underwater shield tunnel

The opening of segment joints is vital to structural stability and leakage prevention for the underwater shield tunnels and it is governed by many environmental variables, e.g., water pressure, earth pressure, and temperature. Identifying the key variables and selecting a reliable predictive model for the joint opening are challenging but important tasks but they are usually done according to subjective judgments. To develop a reliable predictive model, the Nanjing Yangtze River tunnel was introduced as a case study for field monitoring and theoretical analysis. First, the structural health monitoring system in this shield tunnel was used to construct a database of measurements including joint opening, water pressure, earth pressure, and temperatures in the escape layer and inside the segment. Then, Bayesian model class selection was conducted for the variable selection and model construction. Consequently, the water pressure and temperature inside the segment were identified as the input variables for joint opening evaluation. This two-parameter model was validated by using the measurement data. It is useful for the evaluation of similar projects in tunneling engineering.

Estimation and Selection for High-Order Markov Chains with Bayesian Mixture Transition Distribution Models

We develop a mixture model and diagnostic for Bayesian estimation and selection in high-order, discrete-state Markov chains. Both extend the mixture transition distribution, which constructs a transition probability tensor by aggregating probabilities from a set of single-lag transition matrices, through inclusion of mixture components dependent on multiple lags. We demonstrate two uses for the proposed model: identification of relevant lags through over-specification and shrinkage via priors for sparse probability vectors, and parsimonious approximation of multi-lag dynamics by mixing low-order transition models. The diagnostic yields a general and interpretable mixture decomposition for transition probability tensors estimated by any means. We demonstrate the utility of the model and diagnostic with simulation studies, and further apply the methodology to a data analysis from the high-order Markov chain literature, and to a time series of pink salmon abundance in Alaska, United States. Supplemental files for this article are available online.

Biodiversity scale‐dependence and opposing multi‐level correlations underlie differences among taxonomic, phylogenetic and functional diversity

AbstractAimBiodiversity is a multi‐dimensional property of biological communities that represents different information depending on how it is measured, but how dimensions relate to one another and under what conditions is not well understood. We explore how taxonomic, phylogenetic, and functional diversity can differ in scale‐of‐effect dependence and habitat‐biodiversity relationships, and subsequently how spatial differences among biodiversity dimensions may arise.LocationNebraska, United States.TaxonBirds.MethodsAcross 2016 and 2017, we conducted 2,641 point counts at 781 sites. We modeled the occupancy of 141 species using Bayesian Bernoulli‐Bernoulli hierarchical logistic regressions. We calculated species richness (SR), phylogenetic diversity (PD) and functional diversity (FD) for each site and year based on predicted occupancy, accounting for imperfect detection. Using Bayesian latent indicator scale selection and multivariate modeling, we quantified the spatial scales‐of‐effect that best explained the relationships between environmental characteristics and SR, PD and FD. Additionally, we decomposed the residual between‐site and within‐site biodiversity correlations using our repeated measures design.ResultsAlthough relationships between specific land cover types and SR, PD and FD were qualitatively similar, the spatial scales at which these variables were important in explaining biodiversity differed among dimensions. Between‐site residual biodiversity correlations were negative, yet within‐site biodiversity residual correlations were positive.Main conclusionsOur results demonstrate how spatial differences among biodiversity dimensions may arise from biodiversity‐specific scale‐dependent habitat relationships, low shared environmental correlations and opposing residual correlations between dimensions, suggesting that single‐scale and single‐dimension analyses are not entirely appropriate for quantifying habitat‐biodiversity relationships. After accounting for shared habitat relationships, we found positive within‐site residual correlations between SR, PD and FD, suggesting that habitat change over time influenced all biodiversity dimensions similarly. However, negative between‐site residual correlation among biodiversity dimensions may indicate trade‐offs in achieving maximum biodiversity across multiple biodiversity dimensions at any given location.