Model Selection Framework Research Articles

A model selection framework to quantify microvascular liver function in gadoxetate‐enhanced MRI: Application to healthy liver, diseased tissue, and hepatocellular carcinoma

We introduce a novel, generalized tracer kinetic model selection framework to quantify microvascular characteristics of liver and tumor tissue in gadoxetate-enhanced dynamic contrast-enhanced MRI (DCE-MRI). Our framework includes a hierarchy of nested models, from which physiological parameters are derived in 2 regimes, corresponding to the active transport and free diffusion of gadoxetate. We use simulations to show the sensitivity of model selection and parameter estimation to temporal resolution, time-series duration, and noise. We apply the framework in 8 healthy volunteers (time-series duration up to 24 minutes) and 10 patients with hepatocellular carcinoma (6 minutes). The active transport regime is preferred in 98.6% of voxels in volunteers, 82.1% of patients' non-tumorous liver, and 32.2% of tumor voxels. Interpatient variations correspond to known co-morbidities. Simulations suggest both datasets have sufficient temporal resolution and signal-to-noise ratio, while patient data would be improved by using a time-series duration of at least 12 minutes. In patient data, gadoxetate exhibits different kinetics: (a) between liver and tumor regions and (b) within regions due to liver disease and/or tumor heterogeneity. Our generalized framework selects a physiological interpretation at each voxel, without preselecting a model for each region or duplicating time-consuming optimizations for models with identical functional forms.

A meta-learning based distribution system load forecasting model selection framework

This paper presents a meta-learning based, automatic distribution system load forecasting model selection framework. The framework includes the following processes: feature extraction, candidate model preparation and labeling, offline training, and online model recommendation. Using load forecasting needs and data characteristics as input features, multiple metalearners are used to rank the candidate load forecast models based on their forecasting accuracy. Then, a scoring-voting mechanism is proposed to weights recommendations from each meta-leaner and make the final recommendations. Heterogeneous load forecasting tasks with different temporal and technical requirements at different load aggregation levels are set up to train, validate, and test the performance of the proposed framework. Simulation results demonstrate that the performance of the meta-learning based approach is satisfactory in both seen and unseen forecasting tasks.

Growth of matter perturbations in the extended viscous dark energy models

In this work, we study the extended viscous dark energy models in the context of matter perturbations. To do this, we assume an alternative interpretation of the flat Friedmann–Lemaître–Robertson–Walker Universe, through the nonadditive entropy and the viscous dark energy. We implement the relativistic equations to obtain the growth of matter fluctuations for a smooth version of dark energy. As result, we show that the matter density contrast evolves similarly to the Lambda CDM model in high redshift; however, in late time, it is slightly different from the standard model. Using the latest geometrical and growth rate observational data, we carry out a Bayesian analysis to constrain parameters and compare models. We see that our viscous models are compatible with cosmological probes, and the Lambda CDM recovered with a 1sigma confidence level. The viscous dark energy models relieve the tension of H_0 in 2 sim 3 sigma . Yet, by involving the sigma _8 tension, some models can alleviate it. In the model selection framework, the data discards the extended viscous dark energy models.

Cosmological model selection from standard siren detections by third-generation gravitational wave observatories

The multi-messenger observation of GW170817 enabled the first historic measurement of the Hubble constant via a standard siren , so-called in analogy to standard candles that enabled the measurement of the luminosity distance versus redshift relationship at small redshift. In the next decades, third-generation observatories are expected to detect hundreds to thousand of gravitational wave events from compact binary coalescences with potentially a joint electromagnetic counterpart. In the present work, we show how future standard siren detections can be used within the framework of Bayesian model selection to discriminate between cosmological models differing by the parameterization of the late-time acceleration. In particular, we found quantitative conditions for the standard Λ CDM model to be favored with respect to other models with varying dark energy content, by reducing the uncertainty in the gravitational determination of the luminosity distance with respect to current expectations.

A novel approach to optimize the regularization and evaluation of dynamical models using a model selection framework

ABSTRACT Orbit superposition models are a non-parametric dynamical modelling technique to determine the mass of a galaxy’s central supermassive black hole (SMBH), its stars, or its dark matter halo. One of the main problems is how to decide which model out of a large pool of trial models based on different assumed mass distributions represents the true structure of an observed galaxy best. We show that the traditional approach to judge models solely by their goodness-of-fit can lead to substantial biases in estimated galaxy properties caused by varying model flexibilities. We demonstrate how the flexibility of the models can be estimated using bootstrap iterations and present a model selection framework that removes these biases by taking the variable flexibility into account in the model evaluation. We extend the model selection approach to optimize the degree of regularization directly from the data. Altogether, this leads to a significant improvement of the constraining power of the modelling technique. We show with simulations that one can reconstruct the mass, anisotropy, and viewing angle of an axisymmetric galaxy with a few per cent accuracy from realistic observational data with fully resolved line-of-sight velocity distributions (LOSVDs). In a first application, we reproduce a photometric estimate of the inclination of the disc galaxy NGC 3368 to within 5° accuracy from kinematic data that cover only a few sphere-of-influence radii around the galaxy’s SMBH. This demonstrates the constraining power that can be achieved with orbit models based on fully resolved LOSVDs and a model selection framework.

Model diversity and the embarrassment of riches

ABSTRACT In a recent special issue dedicated to the work of Dani Rodrik, Grüne-Yanoff and Marchionni [(2018). Modeling model selection in model pluralism. Journal of Economic Methodology, 25(3), 265–275. https://doi.org/10.1080/1350178X.2018.1488572] raise a potentially damning problem for Rodrik's suggestion that progress in economics should be understood and measured laterally, by a continuous expansion of new models. They argue that this could lead to an ‘embarrassment of riches’, i.e. the rapid expansion of our model library to such an extent that we become unable to choose between the available models, and thus needs to be solved to make ‘model pluralism’ viable. Drawing on Veit’s [(2019a). Model pluralism. Philosophy of the Social Sciences, 50(2), 91–114. https://doi.org/10.1177/0048393119894897] ‘model pluralism’ account, this paper argues that model pluralism as a thesis about the relationship between science and nature undermines the very idea of a general model selection framework for policy making.

How Size and Condition Influence Survival and Cause‐Specific Mortality of Female Elk

ABSTRACTThe size of animal populations fluctuates with number of births, rate of immigration, rate of emigration, and number of deaths. For many ungulate populations, adult female survival is the most important factor influencing population growth. Therefore, increased understanding of survival and causes of mortality for adult females is fundamental for conservation and management. The objectives of our study were to quantify survival rates of female elk (Cervus canadensis) and determine cause‐specific mortality. We predicted that hunter harvest would be the leading cause of mortality. Further, we predicted that hunters would harvest animals that were in prime age (2–9 yr) and in better condition than elk predated by mountain lions (Puma concolor). From 2015 to 2017, we captured 376 female elk in central Utah, USA. We assessed body size and condition of captured elk, fitted each animal with a global positioning system‐collar, and determined cause of death when we received mortality signals. We estimated survival using Kaplan‐Meier estimates and Cox proportional hazard models within an Akaike's Information Criterion model selection framework to identify covariates that influenced survival. We analyzed differences in size and condition measurements between harvested elk and predated elk using analysis of variance tests. Our best model indicated consistent survival across years; mean survival was 78.3 ± 3.5% (SE) including hunter harvest and 95.5 ± 1.7% without hunter harvest. In decreasing order of importance, elk mortality occurred from hunter harvest (21.2%), mountain lion predation (3.7%), depredation removal (0.5%), automobile collision (0.3%), disease (0.3%), complications during calving (0.3%), and those characterized as undetermined (1.3%). Neck circumference and body length were negatively associated with survival, suggesting that larger animals in good condition had lower survival as a result of hunter harvest. Individuals that died because of cougar predation were smaller and had less loin muscle than the average animal. Hunters removed large, healthy, prime‐aged females, individuals that likely have a greater effect on population growth than elk lost to other predators. If the proportion of larger, healthy females in the population begins to decline, hunting practices may require adjustment because hunters may be removing individuals with the greatest reproductive value. © 2021 The Wildlife Society.

Generic Constraint-based Block Modeling using Constraint Programming

Block modeling has been used extensively in many domains including social science, spatial temporal data analysis and even medical imaging. Original formulations of the problem modeled it as a mixed integer programming problem, but were not scalable. Subsequent work relaxed the discrete optimization requirement, and showed that adding constraints is not straightforward in existing approaches. In this work, we present a new approach based on constraint programming, allowing discrete optimization of block modeling in a manner that is not only scalable, but also allows the easy incorporation of constraints. We introduce a new constraint filtering algorithm that outperforms earlier approaches, in both constrained and unconstrained settings, for an exhaustive search and for a type of local search called Large Neighborhood Search. We show its use in the analysis of real datasets. Finally, we show an application of the CP framework for model selection using the Minimum Description Length principle.

The Winter Habitat Selection of Red Deer (Cervus elaphus) Based on a Multi-Scale Model.

Simple SummaryMost wildlife habitat studies are yet to adopt a multi-scale framework. Our study of the winter habitat selection of red deer (Cervus elaphus) is based on a multi-scale model and shows the importance of taking different scales into account when investigating habitat selection. Our approach captures a wide spectrum of ecological relationships of a population, results in effective conservation planning, and is readily applicable to other species of wildlife. Efficacy of future habitat selection studies will benefit by taking a multi-scale approach. In addition to potentially providing increased explanatory power and predictive capacity, multi-scale habitat models enhance our understanding of the scales at which species respond to their environment, which is critical knowledge required to implement effective conservation and management strategies.Single-scale frameworks are often used to analyze the habitat selections of species. Research on habitat selection can be significantly improved using multi-scale models that enable greater in-depth analyses of the scale dependence between species and specific environmental factors. In this study, the winter habitat selection of red deer in the Gogostaihanwula Nature Reserve, Inner Mongolia, was studied using a multi-scale model. Each selected covariate was included in multi-scale models at their “characteristic scale”, and we used an all subsets approach and model selection framework to assess habitat selection. The results showed that: (1) Univariate logistic regression analysis showed that the response scale of red deer to environmental factors was different among different covariate. The optimal scale of the single covariate was 800–3200 m, slope (SLP), altitude (ELE), and ratio of deciduous broad-leaved forests were 800 m in large scale, except that the farmland ratio was 200 m in fine scale. The optimal scale of road density and grassland ratio is both 1600 m, and the optimal scale of net forest production capacity is 3200 m; (2) distance to forest edges, distance to cement roads, distance to villages, altitude, distance to all road, and slope of the region were the most important factors affecting winter habitat selection. The outcomes of this study indicate that future studies on the effectiveness of habitat selections will benefit from multi-scale models. In addition to increasing interpretive and predictive capabilities, multi-scale habitat selection models enhance our understanding of how species respond to their environments and contribute to the formulation of effective conservation and management strategies for ungulata.

Ecological and historical legacies on global diversity gradients in marine elapid snakes

AbstractGlobal diversity gradients have been extensively investigated for several biological groups. However, little is known whether the diversity drivers of clades that underwent major environmental transition (e.g., from land to sea) are equivalent across these different environmental settings. Here, we ask if the pattern of diversity of marine elapid snakes is determined by factors analogous to those previously found for terrestrial lineages. Through a model selection framework, we compare the effect of factors that represent five ecological and historical hypotheses. We found that both ecological and historical factors play significant roles, but habitat structure, which can be linked to historical climatic changes, was better supported. This result agrees with that previously found for terrestrial elapids, despite the different environmental pressures in terrestrial and marine contexts. Our findings suggest an equivalence of the underlying process of diversity gradient within the same lineage from land and sea, irrespective of the different physiological, spatial and historical constraints.

Community health worker-based mobile health (mHealth) approaches for improving management and caregiver knowledge of common childhood infections: A systematic review.

BackgroundChildren in lower middle-income countries (LMICs) are more at risk of dying, than those in High Income Countries (HICs), due to highly prevalent deadly yet preventable childhood infections. Alongside concerns about the incidence of these infections, there has been a renewed interest in involving community health workers (CHWs) in various public health programs. However, as CHWs are increasingly asked to take on different tasks there is a risk that their workload may become unmanageable. One solution to help reduce this burden is the use of mobile health (mHealth) technology in the community through behaviour change. Considering there are various CHWs based mHealth approaches on illness management and education, therefore, we aimed to appraise the available literature on effectiveness of these mHealth approaches for caregivers to improve knowledge and management about common under-five childhood infections with respect to behaviour change.MethodsWe searched six databases between October to December 2019 using subject heading (Mesh) and free text terms in title or abstract in US English. We included multiple study types of children under-five or their caregivers who have been counselled, educated, or provided any health care service by CHWs for any common paediatric infectious diseases using mHealth. We excluded articles published prior to 1990 and those including mHealth technology not coming under the WHO definition. A data extraction sheet was developed and titles, abstracts, and selected full text were reviewed by two reviewers. Quality assessment was done using JBI tools.ResultsWe included 23 articles involving around 300 000 individuals with eight types of study designs. 20 studies were conducted in Africa, two in Asia, and one in Latin America mainly on pneumonia or respiratory tract infections followed by malaria and diarrhoea in children. The most common types of Health approaches were mobile applications for decision support, text message reminders and use of electronic health record systems. None of the studies employed the use of any behaviour change model or any theoretical framework for selection of models in their studies.ConclusionsCoupling mhealth with CHWs has the potential to benefit communities in improving management of illnesses in children under-five. High quality evidence on impact of such interventions on behaviour is relatively sparse and further studies should be conducted using theoretically informed behaviour change frameworks/models.RegistrationPROPSERO Registration number: CRD42018117679

Inferring the Allelic Series at QTL in Multiparental Populations.

Multiparental populations (MPPs) are experimental populations in which the genome of every individual is a mosaic of known founder haplotypes. These populations are useful for detecting quantitative trait loci (QTL) because tests of association can leverage inferred founder haplotype descent. It is difficult, however, to determine how haplotypes at a locus group into distinct functional alleles, termed the allelic series. The allelic series is important because it provides information about the number of causal variants at a QTL and their combined effects. In this study, we introduce a fully Bayesian model selection framework for inferring the allelic series. This framework accounts for sources of uncertainty found in typical MPPs, including the number and composition of functional alleles. Our prior distribution for the allelic series is based on the Chinese restaurant process, a relative of the Dirichlet process, and we leverage its connection to the coalescent to introduce additional prior information about haplotype relatedness via a phylogenetic tree. We evaluate our approach via simulation and apply it to QTL from two MPPs: the Collaborative Cross (CC) and the Drosophila Synthetic Population Resource (DSPR). We find that, although posterior inference of the exact allelic series is often uncertain, we are able to distinguish biallelic QTL from more complex multiallelic cases. Additionally, our allele-based approach improves haplotype effect estimation when the true number of functional alleles is small. Our method, Tree-Based Inference of Multiallelism via Bayesian Regression (TIMBR), provides new insight into the genetic architecture of QTL in MPPs.

ESTABLISHING A SELECTION FRAMEWORK FOR A LEAN AND VALUE ENGINEERING HYBRID METHODOLOGY

The evolution of continuous improvement methodologies is evident in the formalisation of activities, their spread across disciplines, and the formation of hybrid methodologies. This study focuses on hybrid methodology formation, and specifically on lean and value engineering. It is proposed that hybrid methodologies be implemented using interchangeable use, concurrent use, and integrated use models; however, the literature does not provide distinct model definitions nor a formal framework to guide model selection. This study uses a systematic literature review to establish a model selection framework based on distinct model definitions. The aim of this framework is to support implementation and to create a standardised platform for future development, ultimately leading to improved and more sustainable success in implementation.

Dynamic ordering design for dose finding in drug-combination trials.

Drug-combination studies have become increasingly popular in oncology. One of the critical concerns in phase I drug-combination trials is the uncertainty in toxicity evaluation. Most of the existing phase I designs aim to identify the maximum tolerated dose (MTD) by reducing the two-dimensional searching space to one dimension via a prespecified model or splitting the two-dimensional space into multiple one-dimensional subspaces based on the partially known toxicity order. Nevertheless, both strategies often lead to complicated trials which may either be sensitive to model assumptions or induce longer trial durations due to subtrial split. We develop two versions of dynamic ordering design (DOD) for dose finding in drug-combination trials, where the dose-finding problem is cast in the Bayesian model selection framework. The toxicity order of dose combinations is continuously updated via a two-dimensional pool-adjacent-violators algorithm, and then the dose assignment for each incoming cohort is selected based on the optimal model under the dynamic toxicity order. We conduct extensive simulation studies to evaluate the performance of DOD in comparison with four other commonly used designs under various scenarios. Simulation results show that the two versions of DOD possess competitive performances in terms of correct MTD selection as well as safety, and we apply both versions of DOD to two real oncology trials for illustration.

The effect of gene flow from unsampled demes in landscape genetic analysis.

An assumption of correlative landscape genetic methods is that genetic differentiation at neutral markers arises solely from the degree to which the intervening landscape between individuals or populations resists gene flow. However, this assumption is violated when gene flow occurs into the sampled population from an unsampled, differentiated deme. This may happen when sampling within only a portion of a population's extent or when closely related species hybridize with the sampled population. In both cases, violation of the modelling assumptions has the potential to reduce landscape genetic model selection accuracy and result in poor inferences. We used individual-based population genetic simulations in complex landscapes within a model selection framework to explore the potential confounding effect of gene flow from unsampled demes. We hypothesized that as gene flow from outside the sampling extent increased, model selection accuracy would decrease due to the formation of a hybrid zone where allele frequencies were perturbed in a way that was not correlated with effective distances between sampled individuals. Surprisingly, we found this expectation was unfounded, because the reduced accuracy due to admixture was counteracted by an increase in allelic diversity as alleles spread from the unsampled deme into the sampled population. These new alleles increased the power to detect landscape genetic relationships and even slightly improving model selection accuracy overall. This is a reassuring result, suggesting that sampling the full extent of a population or related species that may hybridize may be unnecessary, as long as other well-established sampling requirements are met.

Simultaneous Dimensionality and Complexity Model Selection for Spectral Graph Clustering

Our problem of interest is to cluster vertices of a graph by identifying underlying community structure. Among various vertex clustering approaches, spectral clustering is one of the most popular methods because it is easy to implement while often outperforming more traditional clustering algorithms. However, there are two inherent model selection problems in spectral clustering, namely estimating both the embedding dimension and number of clusters. This article attempts to address the issue by establishing a novel model selection framework specifically for vertex clustering on graphs under a stochastic block model. The first contribution is a probabilistic model which approximates the distribution of the extended spectral embedding of a graph. The model is constructed based on a theoretical result of asymptotic normality for the informative part of the embedding, and on simulation results providing a conjecture for the limiting behavior of the redundant part of the embedding. The second contribution is a simultaneous model selection framework. In contrast with traditional approaches, our model selection procedure estimates embedding dimension and number of clusters simultaneously. Based on our conjectured distributional model, a theorem on the consistency of the estimates of model parameters is presented, providing support for the utility of our method. Algorithms for our simultaneous model selection (SMS) for vertex clustering are proposed, demonstrating superior performance in simulation experiments. We illustrate our method via application to a collection of brain graphs.

An integrated approach to measure hunting intensity and assess its impacts on mammal populations

Abstract Unsustainable hunting of wildlife is one of the greatest threats to diverse and healthy forests, yet our understanding of hunting activity is limited by our methods of accurately identifying its intensity and distribution. Several methods have been used to quantify hunting in past studies (e.g. interviews, ranger patrols and camera traps). However, none of these alone have been able to produce precise spatiotemporal measures of hunting activity. In this study, we used a new method to detect hunters through passive acoustic monitors and developed an integrated approach to measure hunting activity while simultaneously assessing its impacts on mammal populations using camera traps. We applied a hierarchical community occupancy model that accounted for the imperfect detection of species on data from 45 trap locations, surveyed from January to June 2018, to investigate the impacts of spatial variation and intensity in hunting pressure on mammal species richness and occurrence in four protected areas in southern Belize. We developed spatiotemporally explicit indices of hunting activity separately from camera trap and acoustic monitor data and used a Bayesian model selection framework to identify predictors of site occurrence for individual species and three functional groups: carnivores, herbivores and omnivores. We found that camera traps under‐detected hunting activity in the region by 939% compared to acoustic monitors. Consistent with our predictions, hunting intensity was negatively correlated with site‐level species richness, with an average decrease in richness of 31% across its range of variation. Occurrence patterns for the three functional groups were also negatively associated with hunting intensity. Often the target of hunters, herbivores, displayed the strongest negative response to hunting, while omnivores were least affected. Synthesis and applications. Unsustainable hunting of wildlife is a global phenomenon with wide‐ranging implications for ecological communities, especially mammals. Our study highlights mammal sensitivity to increasing hunting pressure at the community and species level and emphasizes the necessity for developing robust tools to accurately monitor hunting activity, while also providing a flexible framework for simultaneously assessing hunting and its impacts on mammal communities.

Bayesian approach to inverse scattering with topological priors

We propose a Bayesian inference framework to estimate uncertainties in inverse scattering problems. Given the observed data, the forward model and their uncertainties, we find the posterior distribution over a finite parameter field representing the objects. To construct the prior distribution we use a topological sensitivity analysis. We demonstrate the approach on the Bayesian solution of 2D inverse problems in light and acoustic holography with synthetic data. Statistical information on objects such as their center location, diameter size, orientation, as well as material properties, are extracted by sampling the posterior distribution. Assuming the number of objects known, comparison of the results obtained by Markov Chain Monte Carlo (MCMC) sampling and by sampling a Gaussian distribution found by linearization about the maximum a posteriori estimate show reasonable agreement. The latter procedure has low computational cost, which makes it an interesting tool for uncertainty studies in 3D. However, MCMC sampling provides a more complete picture of the posterior distribution and yields multi-modal posterior distributions for problems with larger measurement noise. When the number of objects is unknown, we devise a stochastic model selection framework.

METROPOLITAN MAMMALS: UNDERSTANDING THE THREATS INSIDE AN URBAN PROTECTED AREA

Protected areas (PA) are widely recognized as conservation cornerstones. However, we still lack information about PA effectiveness in conserving biodiversity. With the accelerating process of urbanization, urban PAs have gained increasing importance. Thus, evaluating their effectiveness is particularly urgent, especially when located in a biodiversity hotspot. The aim of this study was to describe the large and medium-sized mammalian community within Tijuca National Park (TNP) – an urban PA located in the Atlantic Forest biodiversity hotspot – and to investigate if it is affected by proximity to the park border and by potential threats that occur inside the park such as roads, hiking trails, tourism infrastructure, domestic dogs and hunting. A camera-trap survey was conducted in a grid with 42 sampling units from April to September 2016. Each sampling unit was categorized as park border or interior and according to the aforementioned potential threats. A model selection framework was employed to evaluate the effects of threats on species richness and abundance. Our findings suggest that TNP harbors an impoverished fauna of large and medium-sized mammals when compared with larger and more well-preserved Atlantic Forest PAs. Furthermore, our results highlight that the mammalian abundance was affected by edge effects, presence of roads and tourism infrastructure and abundance of domestic dogs. We highlight some management actions to properly control and minimize those threats and increase the effectiveness of TNP in conserving biodiversity

Understanding the relationship between climatic niches and dispersal through the lens of bat wing morphology

AbstractClimate, an important dimension within the classical Hutchinsonian niche concept, represents abiotic variables that influence species distributions. As dispersal allows species to move and colonize new regions with distinct climates, the role of dispersal traits in determining the occupation of particular climatic conditions is hypothesized to be important in climatic niche determination. I test this hypothesis by investigating the effects of interspecific variation in bat dispersal abilities due to wing morphology on species climatic niches. For 74 phyllostomid bat species, I collected their niche positions (i.e. mean value of occupied climatic conditions) along the climatic space as defined by phylogenetic principal components, as well as wing variables related to flight performance and aerodynamics. To test how wing variables are related to species positions along climatic niche axes, I used a framework of phylogenetic regressions and model selection by information theory. I found that wing morphology has an important and positive influence on bat species niche positions in the climatic space. Species with increased relative wing loading have better flight performance which likely allowed them to overcome geographical barriers to dispersal in order to colonize areas with distinct climatic regimes, thus influencing their climatic niche configurations. My results reconcile the morphological role bat wings have on local habitat use with their effects on dispersal and their importance in shaping climatic niches at broad scales. Overall, my findings reinforce the roles of dispersal and complementary life‐history traits in determining species climatic niches. These results imply that traits linked to tracking suitable habitats may be important for adapting to climate change.