Flexible Modeling Framework Research Articles

A unifying switching regime regression framework with applications in health economics

Motivated by three health economics-related case studies, we propose a unifying and flexible regression modeling framework that involves regime switching. The proposal can handle the peculiar distributional shapes of the considered outcomes via a vast range of marginal distributions, allows for a wide variety of copula dependence structures and permits to specify all model parameters (including the dependence parameters) as flexible functions of covariate effects. The algorithm is based on a computationally efficient and stable penalized maximum likelihood estimation approach. The proposed modeling framework is employed in three applications in health economics, that use data from the Medical Expenditure Panel Survey, where novel patterns are uncovered. The framework has been incorporated in the R package GJRM, hence allowing users to fit the desired model(s) and produce easy-to-interpret numerical and visual summaries.

Bayesian transformation model for spatial partly interval-censored data

The transformation model with partly interval-censored data offers a highly flexible modeling framework that can simultaneously support multiple common survival models and a wide variety of censored data types. However, the real data may contain unexplained heterogeneity that cannot be entirely explained by covariates and may be brought on by a variety of unmeasured regional characteristics. Due to this, we introduce the conditionally autoregressive prior into the transformation model with partly interval-censored data and take the spatial frailty into account. An efficient Markov chain Monte Carlo method is proposed to handle the posterior sampling and model inference. The approach is simple to use and does not include any challenging Metropolis steps owing to four-stage data augmentation. Through several simulations, the suggested method's empirical performance is assessed and then the method is used in a leukemia study.

Mixed Poisson process with Stacy mixing variable

Stacy distribution defined for the first time in 1961 provides a flexible framework for modeling of a wide range of real-life behaviors. It appears under different names in the scientific literature and contains many useful particular cases. Homogeneous Poisson processes are appropriate apriori models for the number of renewals up to a given time t > 0. This article mixes them and considers a mixed Poisson process with Stacy mixing variable. We call it a Poisson-Stacy process. The resulting counting process is one of the Generalized Negative Binomial processes, and the distribution of its time-intersections are very-well investigated in the scientific literature. Here we define and investigate their joint probability distributions. Then, the corresponding mixed renewal process is investigated and Exp-Stacy and Erlang-Stacy distributions are defined and partially studied. The article finishes with a simulation study of these stochastic processes. Some plots of the probability density functions, probability mass functions, mean square regressions, and sample paths are drawn together with the corresponding code for the simulations.

Bayesian size-and-shape regression modelling

Building on Dryden et al. (2021), this note presents the Bayesian estimation of a regression model for size-and-shape response variables with Gaussian landmarks. Our proposal fits into the framework of Bayesian latent variable models and, potentially, allows for a highly flexible modelling framework.

Pathways for the thermally optimal design and practice of anaerobic digestion in large-scale biogas plants: Heat transfer modeling and energy analysis

A flexible and configuration-dependent heat model framework is developed, validated, and used to perform in-depth energy analyses and to identify the optimal design and operating conditions in wet anaerobic digesters in stirred reactors with multi-membrane gasholders in modern biogas plants. This framework enables a detailed geometrical representation of the gasholder and digester at different scales, such that different gasholder designs from single- to multi-membrane systems with varying configurations can be studied and compared to increase thermal efficiency. The model operates in the dynamic regime and includes longwave infrared surface-to-surface heat exchanges. This includes the impact of dynamic environmental conditions (e.g., ambient air temperature, wind, rain, and solar radiation) on the thermal behavior of the unit. Throughout the parametric analyses, the influences of key operating and design parameters, as well as synergistic features such as the feasibility and efficiency of various waste heat recovery strategies, are investigated on the basis of the thermal performance of the digester. The analysis shows that, based on a 3,000-m3 double-membrane biogas plant with a raw biogas self-consumption of 7.4%, thermal autarky can be achieved using (i) a three-membrane gasholder (−2.1%), (ii) outgoing digestate heat recovery (−3.7%), and (iii) recovery of heat from biogas purification (−1.6%).

Epidemic thresholds and human mobility

A comprehensive view of disease epidemics demands a deep understanding of the complex interplay between human behaviour and infectious diseases. Here, we propose a flexible modelling framework that brings conclusions about the influence of human mobility and disease transmission on early epidemic growth, with applicability in outbreak preparedness. We use random matrix theory to compute an epidemic threshold, equivalent to the basic reproduction number R_{0}, for a SIR metapopulation model. The model includes both systematic and random features of human mobility. Variations in disease transmission rates, mobility modes (i.e. commuting and migration), and connectivity strengths determine the threshold value and whether or not a disease may potentially establish in the population, as well as the local incidence distribution.

Generalized neural closure models with interpretability

Improving the predictive capability and computational cost of dynamical models is often at the heart of augmenting computational physics with machine learning (ML). However, most learning results are limited in interpretability and generalization over different computational grid resolutions, initial and boundary conditions, domain geometries, and physical or problem-specific parameters. In the present study, we simultaneously address all these challenges by developing the novel and versatile methodology of unified neural partial delay differential equations. We augment existing/low-fidelity dynamical models directly in their partial differential equation (PDE) forms with both Markovian and non-Markovian neural network (NN) closure parameterizations. The melding of the existing models with NNs in the continuous spatiotemporal space followed by numerical discretization automatically allows for the desired generalizability. The Markovian term is designed to enable extraction of its analytical form and thus provides interpretability. The non-Markovian terms allow accounting for inherently missing time delays needed to represent the real world. Our flexible modeling framework provides full autonomy for the design of the unknown closure terms such as using any linear-, shallow-, or deep-NN architectures, selecting the span of the input function libraries, and using either or both Markovian and non-Markovian closure terms, all in accord with prior knowledge. We obtain adjoint PDEs in the continuous form, thus enabling direct implementation across differentiable and non-differentiable computational physics codes, different ML frameworks, and treatment of nonuniformly-spaced spatiotemporal training data. We demonstrate the new generalized neural closure models (gnCMs) framework using four sets of experiments based on advecting nonlinear waves, shocks, and ocean acidification models. Our learned gnCMs discover missing physics, find leading numerical error terms, discriminate among candidate functional forms in an interpretable fashion, achieve generalization, and compensate for the lack of complexity in simpler models. Finally, we analyze the computational advantages of our new framework.

Habitat Heterogeneity, Environmental Feedbacks, and Species Coexistence across Timescales.

AbstractClassic ecological theory explains species coexistence in variable environments. While spatial variation is often treated as an intrinsic feature of a landscape, it may be shaped and even generated by the resident community. All species modify their local environment to some extent, driving changes that can feed back to affect the composition and coexistence of the community, potentially over timescales very different from population dynamics. We introduce a simple nested modeling framework for community dynamics in heterogeneous environments, including the possible evolution of heterogeneity over time due to community-environment feedbacks. We use this model to derive analytical conditions for species coexistence in environments where heterogeneity is either fixed or shaped by feedbacks. Among other results, our approach reveals how dispersal and environmental specialization interact to shape realized patterns of habitat association and demonstrates that environmental feedbacks can tune landscape conditions to allow the stable coexistence of any number of species. Our flexible modeling framework helps explain feedback dynamics that arise in a wide range of ecosystems and offers a generic platform for exploring the interplay between species and landscape diversity.

A Coupled Wildfire-Emission and Dispersion Framework for Probabilistic PM2.5 Estimation

Accurate representation of fire emissions and smoke transport is crucial for current and future wildfire-smoke projections. We present a flexible modeling framework for emissions sourced from the First Street Foundation Wildfire Model (FSF-WFM) to provide a national map for near-surface smoke conditions exceeding the threshold for unhealthy concentrations of particulate matter at or less than 2.5 µm, or PM2.5. Smoke yield from simulated fires is converted to emissions transported by the National Oceanic and Atmospheric Administration’s HYSPLIT model. We present a strategy for sampling from a simulation of ~65 million individual fires, to depict the occurrence of “unhealthy smoke days” defined as 24-h average PM2.5 concentration greater than 35.4 µg/m3 from HYSPLIT. The comparison with historical smoke simulations finds reasonable agreement using only a small subset of simulated fires. The total amount of PM2.5 mass-released threshold of 1015 µg was found to be effective for simulating the occurrence of unhealthy days without significant computational burden.

An ontology-based innovative energy modeling framework for scalable and adaptable building digital twins

Digitalization of buildings and the use of IoT sensing and metering devices are steadily increasing, offering new opportunities for more autonomous, efficient, and flexible buildings. As part of this transformation and inspired by the added value demonstrated in other domains, the concept of a building digital twin that can monitor, simulate, manage, and optimize building operation has received increased interest. To aid such digital twin implementations, accurate and adaptable simulation models are required, which can effectively integrate and utilize the available data. However, traditional building modeling and digital practices, such as Building Information Modeling and white-box modeling tools, are not easily compatible with these requirements. This work presents an innovative and flexible energy modeling framework based on the SAREF ontology. With a basis in the SAREF4BLDG extension for buildings and the defined classes, different models are presented for a selection of typical systems and devices such as spaces, space heaters, dampers, coils, etc. Using the generic semantics and relations of the SAREF4SYST extension, a method for linking and simulating component models is then presented. A proof-of-concept of the modeling framework is provided, showing its application and feasibility to provide a dynamic simulation of the different systems and devices included in a demonstration case. Finally, a future line of work is identified considering the implementation of the modeling framework in an actual building case study, including integration with actual sensing equipment to demonstrate different digital twin services such as performance monitoring, strategy planning, and operational optimization.

SLiM 4: Multispecies Eco-Evolutionary Modeling.

AbstractThe SLiM software framework for genetically explicit forward simulation has been widely used in population genetics. However, it has been largely restricted to modeling only a single species, which has limited its broader utility in evolutionary biology. Indeed, to our knowledge no general-purpose, flexible modeling framework exists that provides support for simulating multiple species while also providing other key features, such as explicit genetics and continuous space. The lack of such software has limited our ability to model higher biological levels such as communities, ecosystems, coevolutionary and eco-evolutionary processes, and biodiversity, which is crucial for many purposes, from extending our basic understanding of evolutionary ecology to informing conservation and management decisions. We here announce the release of SLiM 4, which fills this important gap by adding support for multiple species, including ecological interactions between species such as predation, parasitism, and mutualism, and illustrate its new features with examples.

A Flexible Zero-Inflated Poisson-Gamma Model with Application to Microbiome Sequence Count Data

In microbiome studies, it is of interest to use a sample from a population of microbes, such as the gut microbiota community, to estimate the population proportion of these taxa. However, due to biases introduced in sampling and preprocessing steps, these observed taxa abundances may not reflect true taxa abundance patterns in the ecosystem. Repeated measures, including longitudinal study designs, may be potential solutions to mitigate the discrepancy between observed abundances and true underlying abundances. Yet, widely observed zero-inflation and over-dispersion issues can distort downstream statistical analyses aiming to associate taxa abundances with covariates of interest. To this end, we propose a Zero-Inflated Poisson Gamma (ZIPG) model framework to address these aforementioned challenges. From a perspective of measurement errors, we accommodate the discrepancy between observations and truths by decomposing the mean parameter in Poisson regression into a true abundance level and a multiplicative measurement of sampling variability from the microbial ecosystem. Then, we provide a flexible ZIPG model framework by connecting both the mean abundance and the variability of abundances to different covariates, and build valid statistical inference procedures for both parameter estimation and hypothesis testing. Through comprehensive simulation studies and real data applications, the proposed ZIPG method provides significant insights into distinguished differential variability and mean abundance. Supplementary materials for this article are available online.

Towards multifunctional landscapes coupling low carbon feed and bioenergy production with restorative agriculture: Economic deployment potential of grass‐based biorefineries

Grass-based biomass from grasslands can be used as feedstock in green biorefineries (GBs) that produce a range of bio-products. In addition, adjustments made as part of crop rotation to increase areas under temporary grasslands can yield benefits such as carbon sequestration, increased soil productivity, reduced eutrophication, and reduced need for pesticides. A flexible modeling framework is developed to analyze the deployment options for GBs that use grass-clover to produce protein feed and feedstock for bioenergy. The focus was on optimal deployment, considering system configuration and operation, as well as land use changes designed to increase grass-clover cultivation on cropland. A case study involving 17 counties in Sweden showed that deployment of GB systems could support biomethane and protein feed production corresponding to 5%–60% and 13%–154%, respectively, of biomethane and soybean feed imports to Sweden in Year 2020. This article is protected by copyright. All rights reserved.

Modeling the impact of surveillance activities combined with physical distancing interventions on COVID-19 epidemics at a local level.

Physical distancing and contact tracing are two key components in controlling the COVID-19 epidemics. Understanding their interaction at local level is important for policymakers. We propose a flexible modeling framework to assess the effect of combining contact tracing with different physical distancing strategies. Using scenario tree analyses, we compute the probability of COVID-19 detection using passive surveillance, with and without contact tracing, in metropolitan Barcelona. The estimates of detection probability and the frequency of daily social contacts are fitted into an age-structured susceptible-exposed-infectious-recovered compartmental model to simulate the epidemics considering different physical distancing scenarios over a period of 26 weeks. With the original Wuhan strain, the probability of detecting an infected individual without implementing physical distancing would have been 0.465, 0.515, 0.617, and 0.665 in designated age groups (0-14, 15-49, 50-64, and >65), respectively. As the physical distancing measures were reinforced and the disease circulation decreased, the interaction between the two interventions resulted in a reduction of the detection probabilities; however, despite this reduction, active contact tracing and isolation remained an effective supplement to physical distancing. If we relied solely on passive surveillance for diagnosing COVID-19, the model required a minimal 50% (95% credible interval, 39-69%) reduction of daily social contacts to keep the infected population under 5%, as compared to the 36% (95% credible interval, 22-56%) reduction with contact tracing systems. The simulation with the B.1.1.7 and B.1.167.2 strains shows similar results. Our simulations showed that a functioning contact tracing program would reduce the need for physical distancing and mitigate the COVID-19 epidemics.

Improving structure identifiability of hydrological processes by temporal sensitivity with a flexible modeling framework

Structure identifiability of hydrological processes refers to the ability to select appropriate mathematical representation from multiple structural options. Although the influence of hydrological processes on model behavior varies in time, this temporal relative influence has not been used to identify process structures. In this work, a framework is proposed to investigate the impacts of temporal relative influence of hydrological processes on structure identification. The framework consists of four elements: (1) a blended model, where hydrological process options and parameter values are regarded as integer and continuous decision variables, respectively, is built based on the flexible modeling framework Raven; (2) the temporal sensitivities of each process are derived by using Sobol’ sensitivity analysis on a moving window; (3) performance metrics with and without sensitivity-weights, which are derived from the temporal sensitivities, are used to identify process structures, respectively; and (4) models identified by these two methods are compared to evaluate performance. A synthetic experiment is designed to validate the proposed framework. Three case studies are carried out in the Tongtianhe, Xunhe, and Wudinghe basins in China. Results are summarized as follows: (1) sensitivity-weighted performance metrics improve the identifiability of process structures by considering the temporal relative influence of hydrological processes on model behavior; (2) quickflow and soil evaporation processes are the two most identifiable processes, followed by infiltration, baseflow, and snow balance processes; and (3) the model identified by sensitivity-weighted performance metrics is more efficient than that selected by the performance metric without sensitivity-weights. It is recommended that the temporal sensitivity of hydrological processes be used to identify model structures based on flexible modeling frameworks.

Modern applications of cross-classified random effects models in social and behavioral research: Illustration with R package PLmixed.

Cross-classified random effects models (CCREMs) have been developed for appropriately analyzing data with a cross-classified structure. Despite its flexibility and the prevalence of cross-classified data in social and behavioral research, CCREMs have been under-utilized in applied research. In this article, we present CCREMs as a general and flexible modeling framework, and present a wide range of existing models designed for different purposes as special instances of CCREMs. We also introduce several less well-known applications of CCREMs. The flexibility of CCREMs allows these models to be easily extended to address substantive questions. We use the free R package PLmixed to illustrate the estimation of these models, and show how the general language of the CCREM framework can be translated into specific modeling contexts.

Stochastic atomic modeling and optimization with fullrmc

Understanding materials' atomic structure with a high level of confidence and certainty is often regarded as a very arduous and sometimes impossible task, especially for newer, emerging technology materials exhibiting limited long-range order. Nevertheless, information about atomic structural properties is very valuable for materials science and synthesis. For non-crystalline amorphous and nanoscale materials, using conventional structural determination methods is impossible. Reverse Monte Carlo (RMC) modeling is commonly used to derive models of materials from experimental diffraction data. Here, the latest developments in the fullrmc software package are discussed. Despite its name, fullrmc provides a very flexible modeling framework for solving atomic structures with many methods beyond RMC. The stochastic nature of fullrmc allows it to explore all possible dimensions and degrees of freedom for atomic modeling and create statistical solutions to match measurements. Differing versions of fullrmc are provided as open source or for cloud computing access. The latter includes a modern web-based graphical user interface that incorporates advanced computing and structure-building modules and machine-learning-based components. The main features of fullrmc are presented, including constraint types, boundary conditions, density shape functions and the two running modes: stochastic using a Monte Carlo algorithm and optimization using a genetic algorithm. Capabilities include tools for statistical, mesoscopic and nanoscopic approaches, atomic or coarse-grained models, and smart artificial-intelligence-ready loss functions.

Real-time density nowcasts of US inflation: A model combination approach

We develop a flexible modeling framework to produce density nowcasts for US inflation at a trading-day frequency. Our framework (1) combines individual density nowcasts from three classes of parsimonious mixed-frequency models; (2) adopts a novel flexible treatment in the use of the aggregation function; and (3) permits dynamic model averaging via the use of weights that are updated based on learning from past performance. These features provide density nowcasts that can potentially accommodate non-Gaussian properties. We document the competitive properties of the nowcasts generated from our framework using high-frequency real-time data over the period 2000–2015.

Bayesian estimation of the effect of health inequality in disease detection

BackgroundMeasuring health inequality is essential to ensure that everyone has equal accessibility to health care. Studies in the past have continuously presented and showed areas or groups of people affected by various inequality in accessing the health resources and services to help improve this matter. Alongside, disease prevention is as important to minimise the disease burden and improve health and quality of life. These aspects are interlinked and greatly contributes to one’s health.MethodIn this study, the Gini coefficient and Lorenz curve are used to give an indication of the overall health inequality. The impact of this inequality in granular level is demonstrated using Bayesian estimation for disease detection. The Bayesian estimation used a two-component modelling approach that separates the case detection process and incidence rate using a mixed Poisson distribution while capturing underlying spatio-temporal characteristics. Bayesian model averaging is used in conjunction with the two-component modelling approach to improve the accuracy of estimates by incorporating many candidate models into the analysis instead of using fixed component models. This method is applied to an infectious disease, influenza, in Victoria, Australia between 2013 and 2016 and the corresponding primary health care of the state.ResultThere is a relatively equal distribution of health resources and services pertaining to general practitioners (GP) and GP clinics in Victoria, Australia. Roughly 80 percent of the population shares 70 percent of the number of GPs and GP clinics. The Bayesian estimation with model averaging revealed that access difficulty to health services impacts both case detection probability and incidence rate. Minimal differences are recorded in the observed and estimated incidence of influenza cases considering social deprivation factors. In most years, areas in Victoria’s southwest and eastern parts have potential under-reported cases consistent with their relatively lower number of GP or GP clinics.ConclusionThe Bayesian model estimated a slight discrepancy between the estimated incidence and the observed cases of influenza in Victoria, Australia in 2013-2016 period. This is consistent with the relatively equal health resources and services in the state. This finding is beneficial in determining areas with potential under-reported cases and under-served health care. The proposed approach in this study provides insight into the impact of health inequality in disease detection without requiring costly and time-extensive surveys and relying mainly on the data at hand. Furthermore, the application of Bayesian model averaging provided a flexible modelling framework that allows covariates to move between case detection and incidence models.

Modular Assessment of Rainfall–Runoff Models Toolbox (MARRMoT) v2.1: an object-oriented implementation of 47 established hydrological models for improved speed and readability

Abstract. The Modular Assessment of Rainfall–Runoff Models Toolbox (MARRMoT) is a flexible modelling framework reproducing the behaviour of 47 established hydrological models. This toolbox can be used to calibrate and run models in a user-friendly and consistent way and is designed to facilitate the sharing of model code for reproducibility and to support intercomparison between hydrological models. Additionally, it allows users to create or modify models using components of existing ones. We present a new MARRMoT release (v2.1) designed for improved speed and ease of use. While improved computational efficiency was the main driver for this redevelopment, MARRMoT v2.1 also succeeds in drastically reducing the verbosity and repetitiveness of the code, which improves readability and facilitates debugging. The process to create new models or modify existing ones within the toolbox is also simplified in this version, making MARRMoT v2.1 accessible for researchers and practitioners at all levels of expertise. These improvements were achieved by implementing an object-oriented structure and aggregating all common model operations into a single class definition from which all models inherit. The new modelling framework maintains and improves on several good practices built into the original MARRMoT and includes a number of new features such as the possibility of retrieving more output in different formats that simplifies troubleshooting, and a new functionality that simplifies the calibration process. We compare outputs of 36 of the models in the framework to an earlier published analysis and demonstrate that MARRMoT v2.1 is highly consistent with the previous version of MARRMoT (v1.4), while achieving a 3.6-fold improvement in runtime on average. The new version of the toolbox and user manual, including several workflow examples for common application, are available from GitHub (https://github.com/wknoben/MARRMoT, last access: 12 May 2022; https://doi.org/10.5281/zenodo.6484372, Trotter and Knoben, 2022b).