Explicit Statistical Model Research Articles

Areal and phylogenetic dimensions of word order variation in Indo-European languages

Abstract Both areal and phylogenetic affiliation have been discussed as driving factors of the distribution of word order in the languages of the world. However, disentangling the interaction of these two factors is challenging. Here we take Indo-European as a test case. Word order in this family is largely homogeneous both within areas and within branches, which makes it difficult to assess which factor was more important in shaping the present-day distribution. To break out of this impasse we turn to corpus data and explicit statistical modeling. Building on a parallel corpus of movie subtitles, we investigate word order on the sentence level under stable pragmatic conditions. We measure the similarity of word order variation between pairs of languages with an information-theoretic distance metric. Using cluster analysis and variation partitioning methods these distance metrics show that phylogenetic distance predicts more variation than geographical distance, but the most important predictor is the shared fraction where phylogeny and area overlap. We conclude that word order has evolved along both dimensions and cannot be reduced to a single one.

Benthic microbial biogeographic trends in the North Sea are shaped by an interplay of environmental drivers and bottom trawling effort

Microbial composition and diversity in marine sediments are shaped by environmental, biological, and anthropogenic processes operating at different scales. However, our understanding of benthic microbial biogeography remains limited. Here, we used 16S rDNA amplicon sequencing to characterize benthic microbiota in the North Sea from the top centimeter of 339 sediment samples. We utilized spatially explicit statistical models, to disentangle the effects of the different predictors, including bottom trawling intensity, a prevalent industrial fishing practice which heavily impacts benthic ecosystems. Fitted models demonstrate how the geographic interplay of different environmental and anthropogenic drivers shapes the diversity, structure and potential metabolism of benthic microbial communities. Sediment properties were the primary determinants, with diversity increasing with sediment permeability but also with mud content, highlighting different underlying processes. Additionally, diversity and structure varied with total organic matter content, temperature, bottom shear stress and bottom trawling. Changes in diversity associated with bottom trawling intensity were accompanied by shifts in predicted energy metabolism. Specifically, with increasing trawling intensity, we observed a transition toward more aerobic heterotrophic and less denitrifying predicted metabolism. Our findings provide first insights into benthic microbial biogeographic patterns on a large spatial scale and illustrate how anthropogenic activity such as bottom trawling may influence the distribution and abundances of microbes and potential metabolism at macroecological scales.

Learning to Bound: A Generative Cramér-Rao Bound

The Cramér-Rao bound (CRB), a well-known lower bound on the performance of any unbiased parameter estimator, has been used to study a wide variety of problems. However, to obtain the CRB, requires an analytical expression for the likelihood of the measurements given the parameters, or equivalently a precise and explicit statistical model for the data. In many applications, such a model is not available. Instead, this work introduces a novel approach to approximate the CRB using data-driven methods, which removes the requirement for an analytical statistical model. This approach is based on the recent success of deep generative models in modeling complex, high-dimensional distributions. Using a learned normalizing flow model, we model the distribution of the measurements and obtain an approximation of the CRB, which we call Generative Cramér-Rao Bound (GCRB). Numerical experiments on simple problems validate this approach, and experiments on two image processing tasks of image denoising and edge detection with a learned camera noise model demonstrate its power and benefits.

Research on a Regional Landslide Early-Warning Model Based on Machine Learning—A Case Study of Fujian Province, China

China’s landslide disasters are serious, and regional landslide disaster early-warning is one of the important means of disaster prevention and mitigation. The traditional regional landslide disaster early-warning model, however, is limited by the complex landslide induction mechanism, limited data accumulation, and insufficient big data analysis methods, and has problems such as limited early-warning accuracy and insufficient refinement. In this paper, a machine learning method was introduced into the field of regional landslide disaster warning. From the model construction process of training sample-set construction, sample learning and training, model parameter optimization, model preservation, warning output, and so on, a method for constructing a regional landslide early-warning model based on machine learning was systematically proposed. In the sample learning and training, 80% of the training sample-set was used as the training set, and 20% was used as the test set for five-fold cross validation. The Bayesian Optimization algorithm was used to optimize the model parameters, and the accuracy, ROC curve, and AUC value were used to verify the model accuracy and model generalization ability. With China’s Fujian province as an example, based on nine years of geological and meteorological data (2010–2018), geological environment factors, factors of hazard-affected bodies and historical disaster situations, and rainfall-induced factors in four categories, a total of 26 indicators were used as input characteristic parameters. Six machine learning algorithms were adopted to improve model training; the results showed that the Random Forest algorithm performed the best, giving an accuracy of 92.3%, and was the model with the best generalization ability (AUC was 0.955). The second best was the Artificial Neural Network model, with an accuracy of 0.937 and an AUC of 0.935. Next were the Nearest Neighbor model, the Logistic Regression model, and the Support Vector Machine; the poorest results were from the Decision Tree model. Finally, the typical rainfall-type landslide disaster process in Fujian Province was selected as an example to verify the Random Forest algorithm model. The results showed that compared with the early-warning results of the original explicit statistical model, the hit rate of the new model was 6 times, or equal to that of the original model, and the landslide density in the early-warning area of the new model was 1.6–1.7 times that of the original model. Preliminary verification showed that the new model based on the Random Forest method has obvious advantages, a higher hit rate and a smaller warning area, and can achieve more accurate warnings. The follow-up will continue to track the new landslide disaster situation in the study area and carry out model verification and correction.

A Comparison of Three Popular Methods for Handling Missing Data: Complete-Case Analysis, Inverse Probability Weighting, and Multiple Imputation

Missing data are a pervasive problem in data analysis. Three common methods for addressing the problem are (a) complete-case analysis, where only units that are complete on the variables in an analysis are included; (b) weighting, where the complete cases are weighted by the inverse of an estimate of the probability of being complete; and (c) multiple imputation (MI), where missing values of the variables in the analysis are imputed as draws from their predictive distribution under an implicit or explicit statistical model, the imputation process is repeated to create multiple filled-in data sets, and analysis is carried out using simple MI combining rules. This article provides a non-technical discussion of the strengths and weakness of these approaches, and when each of the methods might be adopted over the others. The methods are illustrated on data from the Youth Cohort (Time) Series (YCS) for England, Wales and Scotland, 1984–2002.

The Franson Experiment as an Example of Spontaneous Breaking of Time-Translation Symmetry

We describe an explicit statistical model of local hidden variables that reproduces the predictions of quantum mechanics for the ideal Franson experiment and sheds light on the physical mechanisms that might be involved in the actual experiment. The crux of our model is the spontaneous breaking of time-translation gauge symmetry by the hidden configurations of the pairs of photons locked in time and energy involved in the experiment, which acquire a non-zero geometric phase through certain cyclic transformations.

MB-GAN: Microbiome Simulation via Generative Adversarial Network.

BackgroundTrillions of microbes inhabit the human body and have a profound effect on human health. The recent development of metagenome-wide association studies and other quantitative analysis methods accelerate the discovery of the associations between human microbiome and diseases. To assess the strengths and limitations of these analytical tools, simulating realistic microbiome datasets is critically important. However, simulating the real microbiome data is challenging because it is difficult to model their correlation structure using explicit statistical models.ResultsTo address the challenge of simulating realistic microbiome data, we designed a novel simulation framework termed MB-GAN, by using a generative adversarial network (GAN) and utilizing methodology advancements from the deep learning community. MB-GAN can automatically learn from given microbial abundances and compute simulated abundances that are indistinguishable from them. In practice, MB-GAN showed the following advantages. First, MB-GAN avoids explicit statistical modeling assumptions, and it only requires real datasets as inputs. Second, unlike the traditional GANs, MB-GAN is easily applicable and can converge efficiently.ConclusionsBy applying MB-GAN to a case-control gut microbiome study of 396 samples, we demonstrated that the simulated data and the original data had similar first-order and second-order properties, including sparsity, diversities, and taxa-taxa correlations. These advantages are suitable for further microbiome methodology development where high-fidelity microbiome data are needed.

Quest for the Best Evolutionary Model.

In the early 1980s, DNA sequencing became a routine and the increasing computing power opened the door to reconstruct molecular phylogenies using probabilistic approaches. DNA sequence alignments provided a large number of positions containing phylogenetic information, which could be extracted using explicit statistical models that described the mutation process using appropriate parameters. Consequently, an active quest started for building increasingly improved (more realistic) statistical models of nucleotide substitution. The simplest model assumed that nucleotide frequencies were in equilibrium and one single category of substitutions. Subsequent models allowed either unequal nucleotide frequencies or separate rates for transitions and transversions. The HKY85 model (Hasegawa et al. in J Mol Evol 22:160, 1985) combined elegantly both options into a single model, which became one of the most useful ones and has been the choice in many molecular phylogenetic studies ever since. The use of improved substitution models such as HKY85 allows reconstructing more accurate and reliable phylogenies, which in turn provide robust frameworks for understanding how biological diversity evolved and for performing a wealth of comparative studies in different disciplines such as ecology, biogeography, developmental biology, biochemistry, genomics, epidemiology, and biomedicine.

Estimating effective population size changes from preferentially sampled genetic sequences.

Coalescent theory combined with statistical modeling allows us to estimate effective population size fluctuations from molecular sequences of individuals sampled from a population of interest. When sequences are sampled serially through time and the distribution of the sampling times depends on the effective population size, explicit statistical modeling of sampling times improves population size estimation. Previous work assumed that the genealogy relating sampled sequences is known and modeled sampling times as an inhomogeneous Poisson process with log-intensity equal to a linear function of the log-transformed effective population size. We improve this approach in two ways. First, we extend the method to allow for joint Bayesian estimation of the genealogy, effective population size trajectory, and other model parameters. Next, we improve the sampling time model by incorporating additional sources of information in the form of time-varying covariates. We validate our new modeling framework using a simulation study and apply our new methodology to analyses of population dynamics of seasonal influenza and to the recent Ebola virus outbreak in West Africa.

Patterns of pollen dispersal and mating in a population of the clonal plant Sagittaria latifolia

Abstract Increased plant size is generally expected to have negative consequences for mating by increasing pollen transfer between flowers of the same plant. Such geitonogamous self‐pollination would then reduce sexual fitness through both female and male function. However, recent theoretical work has indicated that when plants grow clonally, the outward expansion of plants caused by clonal growth might have positive effects on siring without substantially increasing rates of self‐pollination. We investigated patterns of pollen dispersal, selfing and siring in a monoecious population of the clonal plant Sagittaria latifolia, in which clones varied in size and the extent of intermingling with other clones. A spatially explicit statistical model based on the inferred pollen dispersal kernel was constructed to examine the mechanisms underlying observed mating patterns. Pollen dispersal typically occurred over distances that exceeded the spatial extent of clones. There was a positive association between clone size (measured as the number of ramets per genet) and the likelihood that clones were intermingled with the shoots of other clones. Together, these patterns of pollen dispersal and clonal intermingling resulted in a weak positive association between clone size and selfing rates and a strong positive association between clone size and outcross siring success. These patterns were replicated in the spatially explicit model, indicating that the intermingling of clones is an important determinant of mating patterns in this population. Synthesis. Our study provides the first examination of the pollen dispersal kernel for a clonal plant. It is the first study providing empirical support for model predictions that potentially negative effects of increased selfing in large clones might be offset by increased siring success. This implies that the negative consequences of becoming large do not necessarily apply to clonal plants.

Spatially explicit statistical modeling of random and systematic land cover transitions in the main grassland plain of Nech Sar National Park, Ethiopia

The main grassland plain of Nech Sar National Park (NSNP) is a federally managed protected area in Ethiopia designated to protect endemic and endangered species. However, like other national parks in Ethiopia, the park has experienced significant land cover change over the past few decades. Indeed, the livelihoods of local populations in such developing countries are entirely dependent upon natural resources and, as a result, both direct and indirect anthropogenic pressures have been placed on natural parks. While previous research has looked at land cover change in the region, these studies have not been spatially explicit and, as a result, knowledge gaps in identifying systematic transitions continue to exist. This study seeks to quantify the spatial extent and land cover change trends in NSNP, identify the strong signal transitions, and identify and quantify the location of determinants of change. To this end, the author classifies panchromatic aerial photographs in 1986, multispectral SPOT imagery in 2005, and Sentinel imagery in 2019. The spatial extent and trends of land cover change analysis between these time periods were conducted. The strong signal transitions were systematically identified and quantified. Then, the basic driving forces of the change were identified. The locations of these transitions were also identified and quantified using the spatially explicit statistical model. The analysis revealed that over the past three decades (1986–2019), nearly 52% of the study area experienced clear landscape change, out of which the net change and swap change attributed to 39% and 13%, respectively. The conversion of woody vegetation to grassland (~ 5%), subsequently grassland-to-open-overgrazed land (28.26%), and restoration of woody vegetation (0.76%) and grassland (0.72%) from riverine forest and open-overgrazed land, respectively, were found to be the fully systematic transitions whereas the rest transitions were recorded either partly systematic or random transitions. The location of these most systematic land cover transitions identified through the spatially explicit statistical modeling showed drivers due to biophysical conditions, accessibility, and urban/market expansions, coupled with successive government policies for biodiversity management, geo-politics, demographic, and socioeconomic factors. These findings provide important insights into biodiversity loss, land degradation, and ecosystem disruption. Therefore, the model for predicted probability generally suggests a 0.75 km and 0.72 km buffers which are likely to protect forest and grassland from conversion to grassland and open-overgrazed land, respectively.

Improved estimation of macroevolutionary rates from fossil data using a Bayesian framework

AbstractThe estimation of origination and extinction rates and their temporal variation is central to understanding diversity patterns and the evolutionary history of clades. The fossil record provides the only direct evidence of extinction and biodiversity changes through time and has long been used to infer the dynamics of diversity changes in deep time. The software PyRate implements a Bayesian framework to analyze fossil occurrence data to estimate the rates of preservation, origination, and extinction while incorporating several sources of uncertainty. Building upon this framework, we present a suite of methodological advances including more complex and realistic models of preservation and the first likelihood-based test to compare the fit across different models. Further, we develop a new reversible jump Markov chain Monte Carlo algorithm to estimate origination and extinction rates and their temporal variation, which provides more reliable results and includes an explicit estimation of the number and temporal placement of statistically significant rate changes. Finally, we implement a new C++ library that speeds up the analyses by orders of magnitude, therefore facilitating the application of the PyRate methods to large data sets. We demonstrate the new functionalities through extensive simulations and with the analysis of a large data set of Cenozoic marine mammals. We compare our analytical framework against two widely used alternative methods to infer origination and extinction rates, revealing that PyRate decisively outperforms them across a range of simulated data sets. Our analyses indicate that explicit statistical model testing, which is often neglected in fossil-based macroevolutionary analyses, is crucial to obtain accurate and robust results.

Detailed modeling of positive selection improves detection of cancer driver genes

Identifying driver genes from somatic mutations is a central problem in cancer biology. Existing methods, however, either lack explicit statistical models, or use models based on simplistic assumptions. Here, we present driverMAPS (Model-based Analysis of Positive Selection), a model-based approach to driver gene identification. This method explicitly models positive selection at the single-base level, as well as highly heterogeneous background mutational processes. In particular, the selection model captures elevated mutation rates in functionally important sites using multiple external annotations, and spatial clustering of mutations. Simulations under realistic evolutionary models demonstrate the increased power of driverMAPS over current approaches. Applying driverMAPS to TCGA data of 20 tumor types, we identified 159 new potential driver genes, including the mRNA methyltransferase METTL3-METTL14. We experimentally validated METTL3 as a tumor suppressor gene in bladder cancer, providing support to the important role mRNA modification plays in tumorigenesis.

Environmental predictors of stream flow in semi-arid watersheds for biological assessments

Abstract The aim of this study is to test a spatially explicit statistical model to identify indicators of natural stream flow using readily available stream, climate and landscape data. Understanding flow behavior of unmonitored streams at different temporal scales using environmental indicators is of great interest considering the logistic constraints of providing comprehensive flow instrumentation for all stream reaches in a region. Our results have applications to assess human impact in watersheds, to study environmental changes in fresh water resources, and in the management of local ecosystem. This study uses classification and regression tree analysis to identify significant explanatory variables for a predictive model of stream flow in semi-arid watersheds of southern California, USA. The study collected 77 variables with 30 years record, for a set of 48 sites, interpolated to create raster files at 30 m spatial resolution. After applying Pearson correlation analyses to eliminate redundant variables, nine variables were found to have strong positive predictive value for estimating stream flow at ungauged sites. Nine prediction rasters portraying spatial variation of stream flow for the study region at three key index months during wet, dry, and average rainfall conditions. The predictive power of the variables was tested and cross validated over a subset of data not included when building the model. Model validation by site at monthly temporal resolution showed mixed results. While some sites where accurately predicted others did not. The comparison of observed vs predicted values by month suggest that this statistically based approach is able to predict the general patterns of stream flow at the regional scale, however it may be inaccurate in estimating actual flow values by month since the models tends to under-predict monthly discharge.

The Most Recent Cascadia Earthquake and Native American Narratives

The Cascadia subduction zone fault lies just off the Pacific coast of the USA and Canada. Although this fault has been seismically inactive over the written history of the Cascadia region, it has the potential to produce catastrophic earthquakes and tsunamis. A variety of dating methods have been used to show that the most recent Cascadia earthquake occurred in 1700. Among these methods is an informal analysis of oral traditions handed down by Native American peoples that appear to refer to a major earthquake in this region. A central difficulty in analyzing these narratives quantitatively is their use of a generation and other qualitative measures of time that have no fixed lengths. Here, these narratives are analyzed under an explicit statistical model of the lengths of these measures. The results raise a question about the previous conclusion that these narratives all refer to the most recent Cascadia earthquake.

Survey of Stochastic-Based Computation Paradigms

Effectively tackling the upcoming “zettabytes” data explosion requires a huge quantum leap in our computing power and energy efficiency. However, with the Moore's law dwindling quickly, the physical limits of CMOS technology make it almost intractable to achieve high energy efficiency if the traditional “deterministic and precise” computing model still dominates. Worse yet, the upcoming data explosion mostly comprises statistics gleaned from uncertain, imperfect real-world environment. As such, the traditional computing means of first principles modeling or explicit statistical modeling will very likely be ineffective to achieve flexibility, autonomy, and human interaction. The bottom line is clear: given where we are headed, the fundamental principle of modern computing-deterministic logic circuits can flawlessly emulate propositional logic deduction governed by Boolean algebra-has to be reexamined, and transformative changes in the foundation of modern computing must be made. This paper surveys some of the most important works on stochastic-based computing. We specifically focus on four important research areas: (1) random number generation, (2) stochastic computing, (3) stochastic electronics, and (4) emerging device technology and its potential application in stochastic computing. All these research works share two distinctive features. First, they embrace and exploit quantum-induced randomness as an invaluable information carrier, not the “villain of correct computation” to be suppressed. Second, the theoretical foundation underpinning most of these works are based on neither the Boolean algebra in digital circuit nor the nonlinear amplifying and filtering in analog circuit. Instead, it tightly brings together the algorithmatic essence of computing and the quantum-induced randomness through the powerful framework of stochastic-based computing transformation.

Themis-ml: A Fairness-Aware Machine Learning Interface for End-To-End Discrimination Discovery and Mitigation

ABSTRACTAs more industries integrate machine learning into socially sensitive decision processes like hiring, loan-approval, and parole-granting, we are at risk of perpetuating historical and contemporary socioeconomic disparities. This is a critical problem because on the one hand, organizations who use but do not understand the discriminatory potential of such systems will facilitate the widening of social disparities under the assumption that algorithms are categorically objective. On the other hand, the responsible use of machine learning can help us measure, understand, and mitigate the implicit historical biases in socially sensitive data by expressing implicit decision-making mental models in terms of explicit statistical models. In this article we specify, implement, and evaluate a “fairness-aware” machine learning interface called themis-ml, which is intended for use by individual data scientists and engineers, academic research teams, or larger product teams who use machine learning in production systems.

Projected warming portends seasonal shifts of stream temperatures in the Crown of the Continent Ecosystem, USA and Canada

Climate warming is expected to increase stream temperatures in mountainous regions of western North America, yet the degree to which future climate change may influence seasonal patterns of stream temperature is uncertain. In this study, a spatially explicit statistical model framework was integrated with empirical stream temperature data (approximately four million bi-hourly recordings) and high-resolution climate and land surface data to estimate monthly stream temperatures and potential change under future climate scenarios in the Crown of the Continent Ecosystem, USA and Canada (72,000 km2). Moderate and extreme warming scenarios forecast increasing stream temperatures during spring, summer, and fall, with the largest increases predicted during summer (July, August, and September). Additionally, thermal regimes characteristic of current August temperatures, the warmest month of the year, may be exceeded during July and September, suggesting an earlier onset and extended duration of warm summer stream temperatures. Models estimate that the largest magnitude of temperature warming relative to current conditions may be observed during the shoulder months of winter (April and November). Summer stream temperature warming is likely to be most pronounced in glacial-fed streams where models predict the largest magnitude (> 50%) of change due to the loss of alpine glaciers. We provide the first broad-scale analysis of seasonal climate effects on spatiotemporal patterns of stream temperature in the Crown of the Continent Ecosystem for better understanding climate change impacts on freshwater habitats and guiding conservation and climate adaptation strategies.

Weighted parsimony outperforms other methods of phylogenetic inference under models appropriate for morphology.

One of the lasting controversies in phylogenetic inference is the degree to which specific evolutionary models should influence the choice of methods. Model-based approaches to phylogenetic inference (likelihood, Bayesian) are defended on the premise that without explicit statistical models there is no science, and parsimony is defended on the grounds that it provides the best rationalization of the data, while refraining from assigning specific probabilities to trees or character-state reconstructions. Authors who favour model-based approaches often focus on the statistical properties of the methods and models themselves, but this is of only limited use in deciding the best method for phylogenetic inference-such decision also requires considering the conditions of evolution that prevail in nature. Another approach is to compare the performance of parsimony and model-based methods in simulations, which traditionally have been used to defend the use of models of evolution for DNA sequences. Some recent papers, however, have promoted the use of model-based approaches to phylogenetic inference for discrete morphological data as well. These papers simulated data under models already known to be unfavourable to parsimony, and modelled morphological evolution as if it evolved just like DNA, with probabilities of change for all characters changing in concert along tree branches. The present paper discusses these issues, showing that under reasonable and less restrictive models of evolution for discrete characters, equally weighted parsimony performs as well or better than model-based methods, and that parsimony under implied weights clearly outperforms all other methods.

A spatially explicit statistical model to quantify nutrient sources, pathways, and delivery at the regional scale

Nutrient loading has been linked to many issues including eutrophication, harmful algal blooms, and decreases in aquatic species diversity. In order to develop mitigation strategies to control nutrient sources, the relative contribution and spatial distribution of nutrient sources must be quantified. Though many watershed nutrient models exist in the literature, there is generally a tradeoff between the scale of the model and the level of detail regarding the individual sources of nutrients and basin transport and fate characteristics. To examine the link between watershed nutrient sources, landscape processes, and in-stream loads in the Lower Peninsula of Michigan, a spatially explicit nutrient loading model was developed. The model uses spatially explicit descriptions of nutrient sources and a novel statistical model describing spatially-explicit nutrient attenuation along transport pathways to predict total nitrogen and phosphorus loads. Observations collected during baseflow and melt conditions from 2010–2012 were used to calibrate and validate the model. The model predicts nutrient loads, provides information on the sources of nutrients within each watershed, and estimates the relative contribution of different sources to the overall nutrient load. The model results indicate that there is a high degree of variability in seasonal nutrient export rates, which can be significantly greater during snow melt conditions than during baseflow. In addition, the model highlights the considerable variability in seasonal pathways and processes that impact nutrient delivery. The model performance compares favorably to other regional scale nutrient models. This work has the potential to provide valuable information to environmental managers regarding how and where to target efforts to reduce nutrient loads in surface water.