Univariate Marginal Distributions Research Articles

Copula Approximate Bayesian Computation Using Distribution Random Forests

Ongoing modern computational advancements continue to make it easier to collect increasingly large and complex datasets, which can often only be realistically analyzed using models defined by intractable likelihood functions. This Stats invited feature article introduces and provides an extensive simulation study of a new approximate Bayesian computation (ABC) framework for estimating the posterior distribution and the maximum likelihood estimate (MLE) of the parameters of models defined by intractable likelihoods, that unifies and extends previous ABC methods proposed separately. This framework, copulaABCdrf, aims to accurately estimate and describe the possibly skewed and high-dimensional posterior distribution by a novel multivariate copula-based meta-t distribution based on univariate marginal posterior distributions that can be accurately estimated by distribution random forests (drf), while performing automatic summary statistics (covariates) selection, based on robustly estimated copula dependence parameters. The copulaABCdrf framework also provides a novel multivariate mode estimator to perform MLE and posterior mode estimation and an optional step to perform model selection from a given set of models using posterior probabilities estimated by drf. The posterior distribution estimation accuracy of the ABC framework is illustrated and compared with previous standard ABC methods through several simulation studies involving low- and high-dimensional models with computable posterior distributions, which are either unimodal, skewed, or multimodal; and exponential random graph and mechanistic network models, each defined by an intractable likelihood from which it is costly to simulate large network datasets. This paper also proposes and studies a new solution to the simulation cost problem in ABC involving the posterior estimation of parameters from datasets simulated from the given model that are smaller compared to the potentially large size of the dataset being analyzed. This proposal is motivated by the fact that, for many models defined by intractable likelihoods, such as the network models when they are applied to analyze massive networks, the repeated simulation of large datasets (networks) for posterior-based parameter estimation can be too computationally costly and vastly slow down or prohibit the use of standard ABC methods. The copulaABCdrf framework and standard ABC methods are further illustrated through analyses of large real-life networks of sizes ranging between 28,000 and 65.6 million nodes (between 3 million and 1.8 billion edges), including a large multilayer network with weighted directed edges. The results of the simulation studies show that, in settings where the true posterior distribution is not highly multimodal, copulaABCdrf usually produced similar point estimates from the posterior distribution for low-dimensional parametric models as previous ABC methods, but the copula-based method can produce more accurate estimates from the posterior distribution for high-dimensional models, and, in both dimensionality cases, usually produced more accurate estimates of univariate marginal posterior distributions of parameters. Also, posterior estimation accuracy was usually improved when pre-selecting the important summary statistics using drf compared to ABC employing no pre-selection of the subset of important summaries. For all ABC methods studied, accurate estimation of a highly multimodal posterior distribution was challenging. In light of the results of all the simulation studies, this article concludes by discussing how the copulaABCdrf framework can be improved for future research.

On joint distribution of bids in some k-th price auction model

In the era of increasingly widespread use of electronic communication tools, constant growth of e-commerce turnover is hardly surprising. One of its mechanisms are online auctions whose turnover, on a global scale, is enormous. This paper proposes a stochastic model of the value of subsequent actual bids appearing in the course of a dynamic internet k-th price auction. On this basis, it was possible to determine unconditional joint probability distributions, as well as univariate marginal distributions of successive bids appearing in the course of the auction type under consideration.

Precise large deviations for a multidimensional risk model with regression dependence structure

AbstractIn this paper, we consider a nonstandard multidimensional risk model, in which the claim sizes $\{\vec{X}_k, k\ge 1\}$ form an independent and identically distributed random vector sequence with dependent components. By assuming that there exists the regression dependence structure between inter-arrival time and the claim-size vectors, we extend the regression dependence to a more practical multidimensional risk model. For the univariate marginal distributions of claim vectors with consistently varying tails, we obtain the precise large deviation formulas for the multidimensional risk model with the regression size-dependent structure.

Multivariate probability analysis of wind-wave actions on offshore wind turbine via copula-based analysis

Floating structures in service encounter extreme natural hazards, such as correlated extreme wind-wave loads. This study aims to provide realistic multivariate models of environmental parameters to accurately describe the statistical characteristics of metoceanic conditions, which are essential for the reliability-based design of marine structures. The multiload analysis method includes three major steps: (1) to identify the best-fit univariate marginal distributions of the wind-wave parameters. Mixed models composed of a set of single-mode parametric distributions were found to be more suitable for the measured variables. (2) To construct joint probability distributions of multiple variables using copula-based models, the performance of various models namely, trivariate metaelliptical copulas, Archimedean copulas, and vine copula models. It was observed that joint distributions can be best modelled using vine copula models with considerable flexibility. The final process involves establishing environmental surfaces and contours based on the inverse first-order reliability method. The most unfavourable combination of extreme design conditions can be defined on the environmental surface which is useful for evaluating the response characteristics of offshore wind turbines. These expressions provide a preliminary assessment of wind-wave loads on offshore structures.

Exploring New Horizons: Advancing Data Analysis in Kidney Patient Infection Rates and UEFA Champions League Scores Using Bivariate Kavya–Manoharan Transformation Family of Distributions

In survival analyses, infections at the catheter insertion site among kidney patients using portable dialysis machines pose a significant concern. Understanding the bivariate infection recurrence process is crucial for healthcare professionals to make informed decisions regarding infection management protocols. This knowledge enables the optimization of treatment strategies, reduction in complications associated with infection recurrence and improvement of patient outcomes. By analyzing the bivariate infection recurrence process in kidney patients undergoing portable dialysis, it becomes possible to predict the probability, timing, risk factors and treatment outcomes of infection recurrences. This information aids in identifying the likelihood of future infections, recognizing high-risk patients in need of close monitoring, and guiding the selection of appropriate treatment approaches. Limited bivariate distribution functions pose challenges in jointly modeling inter-correlated time between recurrences with different univariate marginal distributions. To address this, a Copula-based methodology is presented in this study. The methodology introduces the Kavya–Manoharan transformation family as the lifetime model for experimental units. The new bivariate models accurately measure dependence, demonstrate significant properties, and include special sub-models that leverage exponential, Weibull, and Pareto distributions as baseline distributions. Point and interval estimation techniques, such as maximum likelihood and Bayesian methods, where Bayesian estimation outperforms maximum likelihood estimation, are employed, and bootstrap confidence intervals are calculated. Numerical analysis is performed using the Markov chain Monte Carlo method. The proposed methodology’s applicability is demonstrated through the analysis of two real-world data-sets. The first data-set, focusing on infection and recurrence time in kidney patients, indicates that the Farlie–Gumbel–Morgenstern bivariate Kavya–Manoharan–Weibull (FGMBKM-W) distribution is the best bivariate model to fit the kidney infection data-set. The second data-set, specifically that related to UEFA Champions League Scores, reveals that the Clayton Kavya–Manoharan–Weibull (CBKM-W) distribution is the most suitable bivariate model for fitting the UEFA Champions League Scores. This analysis involves examining the time elapsed since the first goal kicks and the home team’s initial goal.

Variable screening based on Gaussian Centered L-moments

An important challenge in big data is identification of important variables. For this purpose, methods of discovering variables with non-standard univariate marginal distributions are proposed. The conventional moments based summary statistics can be well-adopted, but their sensitivity to outliers can lead to selection based on a few outliers rather than distributional shape such as bimodality. To address this type of non-robustness, the L-moments are considered. Using these in practice, however, has a limitation since they do not take zero values at the Gaussian distributions to which the shape of a marginal distribution is most naturally compared. As a remedy, Gaussian Centered L-moments are proposed, which share advantages of the L-moments, but have zeros at the Gaussian distributions. The strength of Gaussian Centered L-moments over other conventional moments is shown in theoretical and practical aspects such as their performances in screening important genes in cancer genetics data.

Misspecification of copula for one-shot devices under constant stress accelerated life-tests

Copula models have attracted significant attention in the recent literature for modeling multivariate observations. An essential feature of copulas is that they enable us to specify the univariate marginal distributions and their joint behaviors separately. This paper provides asymptotic results for misspecification of copula models, and examines the consequences and detection of misspecification in copula models under constant-stress accelerated life-tests for one-shot devices. The one-shot device is considered as a two-component system. The reliability is an important factor in lifetime data applications, and we focus on the effect of misspecification on the estimation of the reliability of one-shot devices. Moreover, the Akaike information criterion is used as a specification test for copula model validation. A simulation study is carried out to evaluate the effect of misspecification under the Gumbel-Hougaard, the Frank and the Clayton copulas incorporated with Weibull and gamma distributions as marginal distributions, in terms of asymptotic bias, asymptotic relative bias, and root mean square error.

Flexible copula model for integrating correlated multi-omics data from single-cell experiments.

With recent advances in technologies to profile multi-omics data at the single-cell level, integrative multi-omics data analysis has been increasingly popular. It is increasingly common that information such as methylation changes, chromatin accessibility, and gene expression are jointly collected in a single-cell experiment. In biomedical studies, it is often of interest to study the associations between various data types and to examine how these associations might change according to other factors such as cell types and gene regulatory components. However, since each data type usually has a distinct marginal distribution, joint analysis of these changes of associations using multi-omics data is statistically challenging. In this paper, we propose a flexible copula-based framework to model covariate-dependent correlation structures independent of their marginals. In addition, the proposed approach could jointly combine a wide variety of univariate marginal distributions, either discrete or continuous, including the class of zero-inflated distributions. The performance of the proposed framework is demonstrated through a series of simulation studies. Finally, it is applied to a set of experimental data to investigate the dynamic relationship between single-cell RNA sequencing, chromatin accessibility, and DNA methylation at different germ layers during mousegastrulation.

Representation of the infimum and supremum of a family of multivariate distribution functions

Sklar's theorem represents a single multivariate distribution function through its univariate marginal distributions and a copula. However, it fails in general when it comes to the representation of the envelopes of a family F of multivariate distribution functions (i.e. its point-wise infimum and supremum), even if we allow more general representing functions than copulas. In this paper we develop an alternative representation which describes the envelopes of a family F in terms of the copulas that correspond to the members of F. Our representation is reminiscent of Sklar's representation but is based on the idea of corner patches of copulas. We prove a series of four representation theorems; (1) a general one, which holds even if the envelopes of the family do not possess a Sklar type representation, (2) a version tailored for the case when the envelopes do possess a Sklar type representation, (3) a theorem for families of continuous distribution functions, and (4) a variant appropriate for modeling dependencies of absolutely continuous random variables. In addition, we demonstrate how our results can be applied to give a Sklar type theorem in the imprecise probability setting, which offers a new approach to multivariate probability boxes.

Estimation in copula models with two-piece skewed margins using the inference for margins method

Copulas provide a versatile tool in the modelling of multivariate distributions. With an increased awareness for possible asymmetry in data, skewed copulas in combination with classical margins have been employed to appropriately model these data. The reverse, skewed margins with a (classical) copula has also been considered, but mainly with classical skew-symmetrical margins. An alternative approach is to rely on a large family of asymmetric two-piece distributions for the univariate marginal distributions. Together with any copula this family of asymmetric univariate distributions provides a powerful tool for skewed multivariate distributions. Maximum likelihood estimation of all parameters involved is discussed. A key step in achieving statistical inference results is an extension of the theory available for generalized method of moments, under non-standard conditions. This together with the inference results for the family of univariate distributions, allows to establish consistency and asymptotic normality of the estimators obtained through the method of ‘inference functions for margins’. The theoretical results are complemented by a simulation study and the practical use of the method is demonstrated on real data examples.

Precise large deviations in a bidimensional risk model with arbitrary dependence between claim-size vectors and waiting times

Consider a bidimensional risk model in which an insurer simultaneously confronts two types of claims sharing a common non-stationary arrival process, and the claim-sizes {X→k;k≥1} form a sequence of i.i.d. random vectors with nonnegative components being dependent on each other. Supposing that the univariate marginal distributions of the claim-size vectors have dominatedly varying tails, precise large deviations for the aggregate amount of claims are obtained, by allowing that the claim-size vectors and claim inter-arrival (waiting) times are arbitrarily dependent.

On New Types of Multivariate Trigonometric Copulas

Copulas are useful functions for modeling multivariate distributions through their univariate marginal distributions and dependence structures. They have a wide range of applications in all fields of science that deal with multivariate data. While there is a plethora of copulas, those based on trigonometric functions, especially in dimensions greater than two, have received much less attention. They are, however, of interest because of the properties of oscillation and periodicity of the trigonometric functions, which can appear in certain models of correlation of natural phenomena. In order to fill this gap, this paper introduces and investigates two new types of “multivariate trigonometric copulas”. Their main theoretical properties are studied, and some perspectives for applications are sketched for future work. In particular, we show that the proposed copulas are symmetric, not associative, with no orthant dependence, and with copula densities that have wide oscillations, which remains an uncommon property in the field. The expressions of their multivariate Spearman’s rho are also determined. Furthermore, the first type of the proposed copulas has the interesting feature of having a multivariate Spearman’s rho equal to 0 for all of the dimensions. Some graphic evidence supports the findings. Some mathematical formulas involving the product of n trigonometric functions may be of independent interest.

A Bayesian Copula Approach for Flood Analysis

This study aims to provide joint modelling of rainfall characteristics in Peninsular Malaysia using two-dimensional copula. Two commonly regarded as important variables in the field of hydrology, namely rainfall severity and duration were derived using the Standard Precipitation Index (SPI) and their univariate marginal distributions are further identified by fitting into several distributions. The paper uses a Bayesian framework to estimate the parameter values in the marginal and copula model. The approximation of the posterior distribution by random sampling has been done by Monte Carlo Markov Chain (MCMC). Next, the authors compared these findings with those based on the classical procedure. The results indicated that the Bayesian approach can be substantially more reliable in parameter estimation for small samples.

Modelling random vectors of dependent risks with different elliptical components

AbstractIn Finance and Actuarial Science, the multivariate elliptical family of distributions is a famous and well-used model for continuous risks. However, it has an essential shortcoming: all its univariate marginal distributions are the same, up to location and scale transformations. For example, all marginals of the multivariate Student’s t-distribution, an important member of the elliptical family, have the same number of degrees of freedom. We introduce a new approach to generate a multivariate distribution whose marginals are elliptical random variables, while in general, each of the risks has different elliptical distribution, which is important when dealing with insurance and financial data. The proposal is an alternative to the elliptical copula distribution where, in many cases, it is very difficult to calculate its risk measures and risk capital allocation. We study the main characteristics of the proposed model: characteristic and density functions, expectations, covariance matrices and expectation of the linear regression vector. We calculate important risk measures for the introduced distributions, such as the value at risk and tail value at risk, and the risk capital allocation of the aggregated risks.

An Example of Augmenting Regional Sensitivity Analysis Using Machine Learning Software

AbstractRegional sensitivity analysis, RSA, has been widely applied in assessing the parametric sensitivity of environmental and hydrological models, in part because of its inherent simplicity. In that spirit, this paper reports an example of an augmented approach to improve its utility in ranking parameter importance, beyond reliance solely on the univariate marginal distributions, to include parametric interactions. Both a deterministic and a stochastic model of the transmission of dengue, an important mosquito‐borne disease, were used to explore the effect of interactions to parameter importance ranking using random forests, a commonly used method based on decision trees. The importance ranking based on random forests was generally consistent with the ranking computed from earlier methods that only examined marginal distributions, but with increased importance shown by several interacting parameters. In addition, and building on an earlier application of tree‐structured density estimation, recently developed software was used to map the regions of the parameter space supporting good fits to calibration data. These methods were also found useful in revealing the scale dependence of sensitivity analysis as well as providing a means of identifying alternative explanations for the observed behavior of the system that remain consistent with calibration criteria, a phenomenon known as equifinality.

Extra-parametrized extreme value copula : Extension to a spatial framework

Hazard assessment at a regional scale may be performed thanks to a spatial model for maxima that can be obtained by combining the generalized extreme-value (GEV) distribution for the univariate marginal distributions with extreme-value copulas to describe their dependence structure, as justified by the theory of multivariate extreme values. A flexible class of extreme-value copulas, called XGumbel for short, combines two Gumbel copulas with extra-parameters weighting each dimension. In a multisite study, the XGumbel copula quickly becomes over-parametrized. In addition, interpolation to ungauged locations is not easily achieved. We develop an extension of the XGumbel copula to the spatial framework by defining the extra-parameters as a mapping shaped as a disk. The inference of the Spatialized XGumbel copula is performed thanks to an Approximate Bayesian Computation (ABC) scheme with summary statistics based on upper tail dependence coefficients. The GEV parameters are estimated with a spatial regression model built with a vector generalized linear model. We evaluate and compare this spatial model with the Brown–Resnick process on annual maxima of daily precipitation totals at 177 gauged stations in the French Mediterranean over a 57 year period. Our analyses show that the ABC scheme yields, except in one instance, interpretable parameters. In addition, the Spatialized XGumbel copula is able to reproduce reasonably well the non-stationarity present in our case study.

A nonparametric copula distribution framework for bivariate joint distribution analysis of flood characteristics for the Kelantan River basin in Malaysia

The joint distribution analysis of multidimensional flood characteristics i.e., flood peak flow, volume and duration, often facilitates a comprehensive understanding in the hydrologic risk assessments. Copula-based methodology are frequently incorporated via parametric approach to model dependence structure of parametric based univariate marginal distributions. But, if the targeted copulas and univariate marginal distributions belongs to some specific parametric families, it might be problematic, if the underlying assumption are violated. Also, no universal rules and literatures are imposed to model any hydrologic vectors and their joint dependence structure through any fixed or pre-defined distributions. In this literature, a nonparametric copula simulation are incorporated and applied as a case study for 50 years annual maximum flood samples of the Kelantan River basin at the Gulliemard bridge station in Malaysia. In this study, a combination of both parametric and nonparametric marginal distribution separately conjoined by a nonparametric copulas framework, which is based on the Beta kernel function. The Beta kernel copula function are incorporated to estimate bivariate copula density which further used to derived joint cumulative density of flood peak-volume, volume-duration and peak-duration pairs and their associated joint as well as conditional return periods.

A Bayesian Framework to Identify Random Parameter Fields Based on the Copula Theorem and Gaussian Fields: Application to Polycrystalline Materials

AbstractFor many models of solids, we frequently assume that the material parameters do not vary in space nor that they vary from one product realization to another. If the length scale of the application approaches the length scale of the microstructure however, spatially fluctuating parameter fields (which vary from one realization of the field to another) can be incorporated to make the model capture the stochasticity of the underlying micro-structure. Randomly fluctuating parameter fields are often described as Gaussian fields. Gaussian fields, however, assume that the probability density function of a material parameter at a given location is a univariate Gaussian distribution. This entails for instance that negative parameter values can be realized, whereas most material parameters have physical bounds (e.g., Young’s modulus cannot be negative). In this contribution, randomly fluctuating parameter fields are therefore described using the copula theorem and Gaussian fields, which allow different types of univariate marginal distributions to be incorporated but with the same correlation structure as Gaussian fields. It is convenient to keep the Gaussian correlation structure, as it allows us to draw samples from Gaussian fields and transform them into the new random fields. The benefit of this approach is that any type of univariate marginal distribution can be incorporated. If the selected univariate marginal distribution has bounds, unphysical material parameter values will never be realized. We then use Bayesian inference to identify the distribution parameters (which govern the random field). Bayesian inference regards the parameters that are to be identified as random variables and requires a user-defined prior distribution of the parameters to which the observations are inferred. For homogenized Young’s modulus of a columnar polycrystalline material of interest in this study, the results show that with a relatively wide prior (i.e., a prior distribution without strong assumptions), a single specimen is sufficient to accurately recover the distribution parameter values.

A two-sample test for the equality of univariate marginal distributions for high-dimensional data

A recurring theme in modern statistics is dealing with high-dimensional data whose main feature is a large number, p, of variables but a small sample size. In this context our aim is to address the problem of testing the null hypothesis that the marginal distributions of p variables are the same for two groups. We propose a test statistic motivated by the simple idea of comparing, for each of the p variables, the empirical characteristic functions computed from the two samples. The asymptotic normality of the test statistic is derived under mixing conditions. In our asymptotic analysis the number of variables tends to infinity, while the size of individual samples remains fixed. In order to obtain a practical test several estimators of the variance are proposed, leading to three somewhat different versions of the test. An alternative global test based on the P-values derived from permutation tests is also proposed. A simulation study to investigate the finite sample properties of the proposed tests is carried out, and a practical illustration involving microarray data is provided.

A generalization of Basu-Dhar’s bivariate geometric distribution to the trivariate case

In this paper, it is introduced a new parametric distribution to be used in multivariate lifetime data analysis as an alternative for the use of some existing multivariate parametric models as the popular multivariate normal distribution which is the most widely used model assumed in the analysis of continuous multivariate data. Although the normal multivariate distribution has univariate marginal normal probability distributions and simple interpretations for all their parameters, it may not be well fitted by many data sets, especially in survival data applications, usually considering logarithm transformed data. In many cases the use of parametric multivariate discrete models could be more appropriate for the data analysis. In this paper, it is introduced a generalization of the bivariate Basu-Dhar geometric distribution to a trivariate case applied to count data. Some properties of this trivariate geometric distribution, including its marginal probability distributions, order statistics distributions, the probability generating function and some simulation studies are presented. It is also presented some discussion on an extension of the trivariate case for the multivariate case. Classical and Bayesian inferences are presented assuming censored or uncensored observations. To illustrate the proposed methodology, two applications with real lifetime data are considered as examples.