In actuarial practice, the modeling of total losses tied to a certain policy is a non-trivial task. Traditional parametric models to predict total losses have limitations due to complex distributional features such as extreme skewness, zero inflation, multi-modality, etc., and the lack of explicit solutions for log-normal convolution. In the recent literature, the application of the Dirichlet process mixture for insurance loss has been proposed to eliminate the risk of model misspecification biases; however, the effect of covariates as well as missing covariates in the modeling framework is rarely studied. In this article, we propose novel connections among covariate-dependent Dirichlet process mixture, log-normal convolution, and missing covariate imputation. Assuming an individual loss is log-normally distributed, we develop a log skew-normal Dirichlet process to approximate the log-normal sum. As a generative approach, our framework models the joint of outcome and covariates, which allows to impute missing covariates under the assumption of missingness at random. The performance is assessed by applying our model to several insurance datasets, and the empirical results demonstrate the benefit of our model compared to the existing actuarial models such as the Tweedie-based generalized linear model, generalized additive model, or multivariate adaptive regression spline.

For order-of-addition experiments, the response is affected by the addition order of the experimental materials. Consequently, the main interest focuses on creating a predictive model and an optimal design for optimizing the response. Van Nostrand (1995) proposed the pairwise-order (PWO) model for detecting PWO effects. Under the PWO model, the full PWO design is optimal under various criteria but is often unaffordable because of the large run size. In this paper, we consider the D-, A- and M.S.-optimal fractional PWO designs. We first present some results on information matrices. Then, a flexible and efficient algorithm is given for generating these optimal PWO designs. Numerical simulation shows that the generated design has an appealing efficiency in comparison with the full PWO design, though with only a small fraction of runs. Several comparisons with existing designs illustrate that the generated designs achieve better efficiencies, and the best PWO designs and some selected 100% efficient PWO designs generated by the new algorithm are reported.

Polynomial recurrence bounds for a class of stochastic differential equations with a gradient type drift and an additive Wiener process are studied without Lyapunov functions.

Let {ξ1, ξ2. . . .} be a sequence of independent possibly differently distributed random variables defined on a probability space (Ω, F, P) with distribution functions {F_ξ1., F_ξ2 , . . .}. Let η be a counting random variable independent of sequence {ξ1, ξ2. . . .}. In this paper, we find conditions under which distribution function of randomly stopped sum Sη = ξ1 + ξ2 + . . . + ξη belongs to the class of generalized subexponential distributions.

We give two examples of the appearance of heavy-tailed distributions in applications to social sciences. Among these distributions are the laws of Pareto, Lotka, and some new ones. The examples are illustrated by constructing suitable toy models.

The multivariate Gaussian random fields with matrix-based scaling laws are widely used for inference in statistics and many applied areas. In such contexts interests are often in symmetry and in the rates of change of spatial surfaces in any given direction. This article analyzes the almost sure sample function behavior for operator fractional Brownian motion, including multivariate fractional Brownian motion. We obtain the estimations of small ball probability and the strongly locally nondeterministic for operator fractional Brownian motion in any given direction. Applying these estimates we obtain Chung's laws of the iterated logarithm for spatial surfaces of operator fractional Brownian motion. Our results show that the precise rates of change of spatial surfaces are completely determined by the self-similarity exponent.

In this paper, under the symmetric entropy and the scale squared error loss functions, we consider the maximum likelihood (ML) estimation and Bayesian estimation of the Shannon entropy and Rényi entropy of the two-parameter inverse Weibull distribution. In the ML estimation, the dichotomy is used to solve likelihood equation. In addition, the approximation confidence interval is given by the Delta method. Because the form of estimation results is more complex in the Bayesian estimation, the Lindley approximation method is used to achieve the numerical calculation. Finally, Monte Carlo simulations and a real data set are used to illustrate the results derived. By comparing the mean square error between the estimated value and the real value, it can be found that the performance of ML estimation of Shannon entropy is better than that of Bayesian estimation, and there is no significant difference between the performance of ML estimation of Rényi entropy and that of Bayesian estimation.

In the first part of this article, we discuss and generalize the complete convergence introduced by Hsu and Robbins (1947) to the r-complete convergence introduced by Tartakovsky (1998). We also establish its relation to the r-quick convergence first introduced by Strassen (1967) and extensively studied by Lai (1976). Our work is motivated by various statistical problems, mostly in sequential analysis. As we show in the second part, generalizing and studying these convergence modes is important not only in probability theory but also to solve challenging statistical problems in hypothesis testing and changepoint detection for general stochastic non-i.i.d. models.

The article proposes a method for determining the p-value of correlations in compositional data, i.e., those data that arise as a result of dividing the original values by their sum. Data organized in this way are typical for many fields of knowledge, but there is still no consensus on methods for interpreting correlations in such data. In a space closed by normalizing quantity, correlation coefficients behave differently than under normal conditions: their probabilities of occurrence do not coincide with those inherent in the standard scale of estimates. In the tens of the new millennium, almost all newly emerging methods for estimating correlation in compositional data began to require mandatory log-transformation of the variable values. In the method proposed here there are no log-transformations. We return to the early stages of attempting to solve the problem and rely on negative shifts in correlations in the multinomial distribution. In modeling the data, we use a hybrid method that combines the hypergeometric distribution with the distribution of any other law. During our work on the calculation method, we found that the number of degrees of freedom in compositional data measures discretely only when all the normalizing sums are equal and that it decreases when the sums are not equal, becoming a continuously varying quantity. Estimation of the number of degrees of freedom and the strength of its influence on the magnitude of the shift in the distribution of correlation coefficients is the basis of the proposed method.

L\'evy walks represent important modeling tools for variety of real life processes. Their natural scaling limits are known to be described by the so-called material fractional derivatives. So far these scaling limits were derived for spatially homogeneous walks, where Fourier and Laplace transforms represent natural tools of analysis. Here we derive the corresponding limiting equations in the case of position depending times and velocities of walks, where Fourier transforms cannot be effectively applied. In fact, we derive three different limits (specified by the way the process is stopped at an attempt to cross the boundary), leading to three different multi-dimensional versions of Caputo-Dzherbashian derivatives, which correspond to different boundary conditions for the generators of the related Feller semigroups and processes. Some other extensions and generalisations are analysed.

In this paper, we will extend the expected value of the function w.r.t the uniform probability measure on sets measurable in the Caratheodory sense to be finite for a larger class of functions, since the set of all measurable functions with infinite or undefined expected values forms a prevalent subset of the set of all measurable functions, which means "almost all" measurable functions have infinite or undefined expected values. Before we define the specific problem in section 2, we will outline some preliminary definitions. We'll then define the specific problem (along with a partial solution in section 3) to visualize the complete solution. Along the way, we will ask a series of questions to clarify our understanding of the paper.

Quantile regression model is widely used in variable relationship research of general size data, due to strong robustness and more comprehensive description of the response variables' characteristics. With the increase of data size and data dimension, there have been some studies on high-dimensional quantile regression under the classical statistical framework, including higher-efficient frequency perspective, which is however at cost of randomness quantification, or lower-efficient Bayesian method based on MCMC sampling. To overcome these problems, we propose the high-dimensional quantile regression with Spike-and-Slab Lasso penalty based on variational Bayesian (VBSSLQR), which can not only improve the computational efficiency but also measure the randomness via variational distributions. The simulation studies and real data analysis illustrate that the proposed VBSSLQR method is superior to or equivalent to other quantile and non-quantile regression methods (including Bayesian and non-Bayesian methods), and its efficiency is higher than any other method.

This study considers the change-point test problem for time series based on the self-normalization ratio statistic test, which is constructed using residuals obtained from a support vector regression (SVR)-autoregressive moving average (ARMA) model. Under the null hypothesis, the series is a stationary process, and our test statistic converges to a non-degenerate distribution. Under the alternative hypothesis, there are change-points in the time series, and the self-normalization test statistic diverges to infinity. The simulations show that our proposed new test has better finite sample performance than other SVR-based tests in the literature. Finally, we illustrate its usefulness by analyzing two actual data sets.

This article presents six probability distributions from the Gamma family with three parameters, for the flood frequency analysis in hydrology. The choice of the Gamma family of statistical dis-tributions was driven by its frequent use in hydrology. In the Faculty of Hydrotechnics, the im-provement of the estimation of maximum flows and including the methodological bases for the realization of a regionalization study with the linear moments method with the corrected pa-rameters was researched, being an element of novelty. The linear moments method is better than MOM because it avoids the choice of skewness depending on the origin of the flows, practiced in Romania. The L-moments method conforms to the current trend for estimating the parameters of statistical distributions. Observed data from hydrometric stations are of relatively short length, so the statistical parameters that characterize them are of a sample that requires correction. The correction of the statistical parameters is proposed, using the method of least squares based on the inverse functions of the statistical distributions expressed with the frequency factor for L-moments. All the necessary elements for their use are presented like, quantile functions, the exact and ap-proximate relations for estimating parameters and frequency factors. A flood frequency analysis case study was carried out for the Ialomita river, to verify the proposed methodology. The per-formance of this distributions is evaluated using Kiling-Gupta and Nash-Sutcliff coefficients.

This article presents the results of the consideration of the Bertrand paradox in terms of geometric probability theory. Although the method and conclusions of this article are straightforward, no equivalent studies were found when reviewing the relevant literature. This could be conditioned by the fact that the hypotheses presented in this research have low intuitive obviousness and, in contrast, could be due to the historically established agreement regarding the issue. This article shows that out of three classical solutions to the problem described by Bertrand, two methods are inconsistent with the claimed relative objectivity. Although the remaining solution (1/4) seems to be the most correct, we cannot claim that we have exhaustively ruled out all aspects that could reduce its adequacy to solve the problem.

Multilevel functional principal components analysis (mFPCA) is used to treat dynamical changes in shape in this article. Results of standard (single-level) FPCA are presented here also as a comparison. Monte Carlo (MC) simulation is used to create univariate data (i.e., a single “outcome” variable) that contains two distinct classes of trajectory with time. MC simulation is also used to create multivariate data of sixteen 2D points that represent (broadly) an eye; this data also has two distinct classes of trajectory (an eye blinking and an eye widening in surprise). This is followed-up by an application of mFPCA and single-level FPCA to “real” data consisting of twelve 3D landmark outlining the mouth that are tracked over all phases of a smile. By consideration of eigenvalues, results for the MC datasets find correctly that variation due to differences in groups between the two classes of trajectories variation are larger than variation within each group. In both cases, differences in standardized component scores between the two groups are observed, as expected. Modes of variation are shown to model the univariate MC data correctly, and good model fits are found for both the “blinking” and “surprised” trajectories for the MC “eye” data. Results for the “smile” data show that the smile trajectory is modelled correctly, namely: the corners of the mouth are drawn backwards and wider during a smile. Furthermore, the first mode of variation at level 1 of the mFPCA model show only subtle and minor changes in mouth shape due to sex, whereas the first mode of variation at level 2 of the mFPCA model govern whether the mouth is upturned or downturned. These results are all an excellent test of mFPCA, which show that mFPCA presents a viable method of modelling dynamical changes in shape.

Building fires are preventable incidents that have proven to be both deadly and costly. Addressing their root causes will lead to safer neighbourhoods for families and businesses to live and operate in. Multiple studies have established the effect of residents’ socioeconomic compositions on an area’s building fire rates; however, the existing model based on the classical stepwise regression procedure has several limitations. This paper aims to construct a more accurate predictive model of building fire rates based on a set of explanatory socioeconomic variables. In building the socioeconomic model, a backward elimination by Robust Final Predictor Error (RFPE) criterion is proposed to enhance the forecasting capability of the model. The proposed method has been implemented on the census data and the fire incident data of the South East Queensland region in Australia. A cross-validation was then conducted to assess the model’s accuracy. In addition, comparative analyses of other elimination criteria, such as p-value, adjusted R-squared, Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC) and predicted residual error sum of squared (PRESS), were conducted. The cross-validation analyses demonstrate that the proposed criterion is a more accurate predictive model based on a couple of goodness-of-fit measures. All in all, the RFPE equation was found to be a suitable criterion for the backward elimination procedure in the socioeconomic modelling of building fires.

We consider a new model of a branching random walk on a multidimensional lattice with continuous time and one source of particle reproduction and death, as well as an infinite number of sources in which, in addition to the walk, only absorption of particles can occur. The asymptotic behavior of the integer moments of both the total number of particles and the number of particles at a lattice point is studied depending on the relationship between the model parameters. In the case of the existence of an isolated positive eigenvalue of the evolution operator of the average number of particles, a limit theorem is obtained on the exponential growth of both the total number of particles and the number of particles at a lattice point.

Bootstrapping is a flexible, powerful and well-established statistical approach to quantify the uncertainty of virtually any point estimate. While multiple versions of bootstrap confidence intervals are already available in Stata, dbs implements the double (iterated) bootstrap. Instead of relying on parametric assumptions such as the non-parametric resampling bootstrap confidence interval does, it is more flexible and derives critical values directly from that data. To do so, multiple methods are available (analytic approach, double resampling, jackknife estimation). In a comparative simulation study it is empirically demonstrated that the strengths of the double bootstrap are particularly evident for small samples (n < 100) when heteroscedasticity is present. While all other approaches result in undercoverage, only the double bootstrap reaches the target coverage level and hence avoids incorrect statistical conclusions. The computational burden is not even necessarily larger than for other bootstrap approaches.

In the era of big data with increasingly complex data structures and ever-larger data scales, matrix-type data are becoming highly valued and their applications in the fields of medicine, industry, education, geography, and astronomy are growing in extent. In recent years, significant progress has been made in the practical use of matrix variable t-distribution finite mixture models for handling data in order to address the issues of multi-subgroup structures and long data tails. In this paper, the expectation-maximization (EM) algorithm with penalized maximum likelihood is proposed to resolve the problem of the unbounded nature of the likelihood function applied to the model by considering the degeneracy of the variance-covariance matrix of this model. Our data were analyzed through simulations and real data, and the results demonstrate that our model is effective in both preventing the likelihood function from being unbounded and in ensuring the accuracy of the estimated parameters of the EM algorithm.

Many papers in recent years have been devoted to estimating the per pixel proportions of three broad classes of materials (e.g. photosynthetic vegetation, non-photosynthetic vegetation and bare soil) using data from multispectral sensors. Many of these papers use estimation methods based on the linear mixture model. Very few of these papers assess the accuracy of their estimators. I show how to produce confidence intervals (CIs) and joint confidence regions (JCRs) for the proportions associated with various linear mixture models. There are two main models, both of which assume that the coefficients in the model are non-negative. The first model assumes that the coefficients sum to 1. The second does not, but uses rescaling of the estimated coefficients to produce estimated proportions. Three variants of these two models are also analysed. JCRs are shown to be particularly informative, because they are typically better at localising the information than CIs are. The methodology is illustrated using examples from Landsat Thematic Mapper data at 1169 locations across Australia, each of which has associated field observations. There is also discussion about the extent to which the methodology can be extended to hyperspectral data.

The opioid crisis in the United States has had devastating effects on communities across the country, leading many states to pass legislation that limits the prescription of opioid medications in an effort to reduce the number of overdose deaths. This study evaluates the impact of two categories of PDMP and Pill Mill regulations on the supply of opioid prescriptions at the level of dispensers and distributors (excluding manufacturers) using ARCOS data. The study uses a difference-in-difference method with a two-way fixed design to analyze the data. The study finds that both of the regulations are associated with reductions in the volume of opioid distribution. However, the study reveals that these regulations may have unintended consequences, such as shifting the distribution of controlled substances to neighboring states. For example, in Tennessee, the implementation of Operational PDMP regulations reduces in-state distribution of opioid drugs by 3.36% (95% CI, 2.37 to 4.3), while the out-of-state distribution to Georgia, which did not have effective PDMP regulations in place, increases by 16.93% (95% CI, 16.42 to 17.44). Our studies emphasize that policymakers should consider the potential for unintended distribution shifts of opioid drugs to neighboring states with laxer regulations as well as varying impacts on different dispenser types.

For each $t\in(-1,1)$, exact values of the least upper bounds $$ H(t)=\sup_{\E X=t,\,\E X^2=1} \frac{\E\abs{X}^3}{\E \abs{X-t}^3},\quad \sup_{\E X=t,\,\E X^2=1} \frac{L_1(X)}{L_1(X-t)} $$ are obtained, where $L_1(X)=\E|X|^3/(\E X^2)^{3/2}$ is the non-cental Lyapunov ratio. It is demonstrated that these values are attained only at two-point distributions. As a corollary, S.\,Shorgin's conjecture is proved that states that the exact value is $$ \sup\frac{L_1(X)}{L_1(X-\E X)}= \frac{\sqrt{17 + 7\sqrt7}}{4} = 1.4899\ldots, $$ where the supremum is taken over all non-degenerate distributions of the random variable $X$ with the finite third moment. Also, in terms of the central Lyapunov ratio $L_1(X-\E X)$, an analog of the Berry--Esseen inequality is proved for Poisson random sums of independent identically distributed random variables with the constant $$ 0.3031\cdot H\left(\frac{\E X}{\sqrt{\E X^2}}\right) \left(1-\frac{(\E X)^2}{\E X^2}\right)^{3/2} \le 0.3031\cdot \frac{\sqrt{17 + 7\sqrt7}}{4}<0.4517.$$ where $\Law(X)$ is the common distribution of the summands.

In recent times, there has been a growing interest in the use of control charts as a feedback process monitoring technique. Particularly, CUSUM schemes have been proven to be efficient for monitoring processes where the magnitude of the sustained mean shifts is small or moderate. However, because of the constant control limits, the detection ability of CUSUM schemes becomes slow at the initial set-up of the process. The fast initial response FIR feature has often been used to enhance it at the process startup. Meanwhile, the dynamism of real-life problems has always encouraged a more sensitive CUSUM scheme capable of detecting process shifts more rapidly. In this paper, an improved CUSUM scheme for monitoring process mean is proposed. This scheme will be substantial for monitoring processes whose observations are obtained at a distant time interval for example hourly, daily or weekly and where the sustained shifts are assumed to be small or moderate. To demonstrate the practical applications of the proposed scheme, we present its real-life application using datasets from a bottling company and petroleum refinery laboratory.

Statistical modeling of lifetime data plays an essential role in a wide range of practical fields, such as health and engineering. There have been a lot of studies done to develop statistical models that can better describe health data than traditional models. For the first time, we pioneer a novel family of continuous probability distributions called the generalized odd beta prime generalized (GOBP-G) family of distributions. The cumulative distribution and probability density functions of the new family are presented. A new generalization of the Weibull distribution called "generalized odd beta prime-Weibull" (GOBPW) is proposed using the pioneered GOBP-G family. The mixture representations of the new distribution are defined and derived. Some formal statistical properties of the GOBPW distribution, such as the moments, moment generating function, incomplete moments, information generating function, entropies, stress-strength function, quantile function, and order statistics, are derived. The estimation of the parameters of the proposed distribution is evaluated using the maximum likelihood estimation approach. Different cancer disease data sets, such as the bladder, head and neck, acute bone, and blood cancers, are used to illustrate the applicability and usefulness of the new model and were compared using several statistical accuracy measures with that of well-established extended Weibull distributions, which are the beta modified Weibull distribution, Kumaraswamy modified Weibull distribution, gamma generalized modified Weibull distribution, gamma log-logistic Weibull distribution, and beta log-logistic Weibull distribution. The results show that the proposed model gives better results than the competitive models. This study could guide the relevant stakeholders in choosing a suitable statistical model for the health data instead of relying on traditional models to enhance decision-making.

Background. COVID-19 efforts were often ineffective in controlling the spread of the pandemic. Identifying ineffective controls during a pandemic is thus vital. Method. Utilizing publicly available data on COVID deaths in the counties of US states, we create an index to capture and interpret ineffectiveness in the efforts to reduce the spread of the pandemic in US counties. This index is based on the Intervened Poisson Distribution (IPD) introduced originally by Shanmugam. Motivation for the research idea occurred while we noticed the data dispersion of the COVID deaths is smaller than the average only in some counties. Under-dispersed data is common in statistical modeling. A novel approach we adapted in this article includes the estimation of an intervention parameter estimated through iterative non-linear optimization. Results. Twenty-five counties in California, Idaho, Minnesota, Mississippi, Montana, Nebraska, North Carolina, North Dakota, Texas, and Utah were found to be ineffective in controlling for fatalities based on the expected probability distribution. A review of the policies enacted in these areas would provide insight into ineffective prevention efforts, and some of these issues are documented in current literature. Conclusion. The IPD index an innovate way to document efficacy of interventions during pandemics. The IPD may identify ineffective efforts prior to statistical models intended to evaluate efficacy of efforts.

Sword-duels are interesting, from a game-theoretical perspective, as they correspond to sudden-death interactions between players (i.e., defeat means removal from the game), and where victory/defeat depend on not just skill, but also luck. Analyzing probabilities of victory streaks, given a certain level of self and others' skill, is thus relevant for any application domain that can be modeled in the same way. This paper takes inspiration from Miyamoto Musashi, famously undefeated for 61 duels, and implements a Markov-chain Monte-Carlo simulation approach to evaluate this scenario. Results suggest that a 61 victory streak can be probabilistically observed when skill level is roughly 6.5 times that of the average duelist, and with 95\% confidence when skill level is roughly 1000 times that of the average duelist. More generally, this paper provides a method for determining chances of victory streaks in game-theoretical sudden-death encounters, when both "skill" and "luck" contribute to the outcome of the encounter. Specific scenarios can be modeled by modifying the utilized Markov chain and adjusting sampled distributions as required.

Water resource management and flood forecasting are crucial societal and financial stakes requiring reliable predictions of flow parameters (depth, velocity), the accuracy of which is often limited by uncertainties in hydrodynamic numerical models. In this study, we assess the effect of two uncertainty sources, namely breach characteristics induced by overtopping and the roughness coefficient, on water elevations and inundation extent. A two-dimensional (2D) hydraulic solver was applied in a Monte Carlo integration framework to a reach of the Loire river (France) including about 300 physical parameters. Inundation hazard maps for different flood scenarios allowed highlighting the impact of the breach development chronology. A special attention was paid to proposing a relevant sensitivity analysis to exam the factors influencing the depth and extent of flooding. The spatial analysis of vulnerability area induced by a levee breach width exhibits that, with increasing the flood discharge, the rise of the parameter influence is accompanied by a more localized spatial effect. This argues for a local analysis to allow a clear understanding of the flood hazard. The physical interpretation, highlighted by a global sensitivity analysis, showed precisely the dependence of the flood simulation on the main factors studied, i.e. the roughness coefficients and the characteristics of the breaches.

Guessing effects frequently occur in testing data in educational or psychological applications. Different item response models have been proposed to handle guessing effects in dichotomous test items. However, it has been pointed out in the literature that the oftenemployed three-parameter logistic model poses implausible assumptions regarding the guessing process. The four-parameter guessing model has been proposed as an alternative to circumvent these conceptual issues. In this article, the four-parameter guessing model is compared with alternative item response models for handling guessing effects through a simulation study and an empirical example. It turned out that model selection for item response models should be rather based on the AIC than the BIC. However, the RMSD item fit statistic used with typical cutoff values was found to be ineffective in detecting misspecified item response models. Furthermore, sufficiently large sample sizes are required for sufficiently precise item parameter estimation. Moreover, it is argued that the criterion of statistical model fit should not be the sole criterion of model choice. The item response model used in operational practice should be valid with respect to the meaning of the ability variable and the underlying model assumptions. In this sense, the four-parameter guessing model could be the model of choice in educational large-scale assessment studies.

Noisy data is always encountered in real applications, such as bioinformatics, neuroimage and remote sensing. Existing methods mainly consider linear or generalized linear errors-in-variables regression, while relatively little attention is paid for the multivariate response case, and how to evaluate the estimation performance under perturbed covariates is still an open question. In this paper, we consider the information-theoretic limitations of estimating a low-rank matrix in the multi-response errors-in-variables regression model. By application of the information theory and statistical techniques on concentration inequalities, the minimax lower bound is provided in terms of the squared Frobenius loss, which recaptures the rate provided under the clean covariate assumption in previous literatures. Hence our result further indicates that though under the more realistic errors-in-variables situation, no more samples are required so as to achieve a rate-optimal estimation.

Directed acyclic graphs (DAGs) are nonparametric causal path diagrams that have substantial utility as principled representations of disease and healthcare pathways, and of the underlying ‘data generating mechanisms’ these pathways involve. As such, DAGs provide a valuable bridge between: the aetiological knowledge, operational insight and professional experience on which clinical training and practice depend; and the more abstract epistemological and analytical considerations required to extract robust statistical insight from health and healthcare data. DAGs are nonetheless vulnerable to imperfect biomedical paradigms, partial clinical knowledge and limited empirical data. DAGs drawn under such circumstances offer limited scope for statistical insight free from cognitive, analytical or inferential bias if: they misrepresent the data generating mechanisms involved; or ignore the important role that omitted variables (whether measured, unmeasured or unacknowledged) might play therein. To address these weaknesses and broaden the appeal and application of DAGs, this chapter provides ten simple steps that educators can use to improve the analytical competence and statistical confidence of the healthcare students, qualified practitioners and experienced researchers they support. These steps use temporal logic to draw DAGs so as to: reduce reliance on uncertain knowledge, incomplete information, flawed assumptions or guesswork; and avoid, mitigate or acknowledge the errors and biases that each of these incur. The chapter comprises an accessible, non-technical overview of the perspective and thoughtfulness required to generate temporally coherent DAGs as objective representations of the probabilistic causal paths involved in context-specific data generating mechanisms. It encourages a focus on those variables operating as potential sources of analytical or inferential bias when estimating the plausible, probabilistic causal relationship between two pre-specified variables; and specifically addresses the challenges posed by: omitted; time-variant; non-asynchronous; and temporally obscure variables. The chapter includes a worked example based on a published clinical study to demonstrate how each of the steps required to generate temporally-informed DAGs can be applied to: critically appraise the analytical decisions made during applied healthcare research; and inform the decisions required when designing, undertaking and analysing primary and secondary, prospective and retrospective research. The appendices include a summary of ten recommendations for improving the reporting and interrogability of DAGs and DAG-informed analyses.

For more than half a century, Manfred Deistler has been contributing to the construction of the rigorous theoretical foundations of the statistical analysis of time series and more general stochastic processes. Half a century of unremitting activity is not easily summarized in a few pages. In this short note, we chose to concentrate on a relatively little-known aspect of Manfred's contribution which nevertheless had quite an impact on the development of one of the most powerful tools of contemporary time series and econometrics: dynamic factor models.

Link prediction is an unbalanced early retrieval problem, whose goal is to prioritize a small cohort of positive links on top of a list largely populated by unlabelled links. Differently from binary classification, here the evaluation focuses on how the predictor prioritizes the positive class because, in practice, a negative class does not exist. Previous studies explained that AUC-ROC is not apt for unbalanced class problems and is misleading for early retrieval problems, therefore standard AUC-ROC is not appropriate for evaluation of link prediction. However, some scholars argue that an AUC-ROC like evaluation accounting for the relative positioning of the few positive links among the vastness of unlabelled links remains a valid concept to pursue. Here we propose the area under the magnified ROC (AUC-mROC), a new measure that adjusts the standard AUC-ROC to work also for unbalanced early retrieval problems such as link prediction.

Location-and-scale transformation of a random variable underpins normal distribution, but it is however fundamentally incorrect for scale estimation such as relative dispersion. In this paper, a parametrized alternative to a normal distribution, called scaloc-normal distribution, is proposed that efficiently works and is fundamentally correct with absolute and relative dispersion estimators. The Monte Carlos simulation experiment was used to generate a total of 600,000 artificial datasets in 600 different simulation scenarios from loc-normal (normal) and scaloc-normal distributions. The absolute and relative dispersion were estimated and compared from the two distributions. The results show that scaloc-normal distribution is a good parametrized alternative to loc-normal distribution, fundamentally correct and efficient with both standard deviation and coefficient of variation. The key statistical advancement from loc-normal to scaloc-normal distribution is its fundamental correctness (i.e., scale-invariant property) with an efficient relative estimator of dispersion (i.e., coefficient of variation). Parametrically, the loc-normal and scaloc-normal distributions are very different, but both have linear transformations.

Numerous approaches address over-fitting in neural networks: by imposing a penalty on the parameters of the network (L1, L2, etc); by changing the network stochastically (drop-out, Gaussian noise, etc.); or by transforming the input data (batch normalization, etc.). In contrast, we aim to ensure that a minimum amount of supporting evidence is present when fitting the model parameters to the training data. This, at the single neuron level, is equivalent to ensuring that both sides of the separating hyperplane (for a standard artificial neuron) have a minimum number of data points — noting that these points need not belong to the same class for the inner layers. We firstly benchmark the results of this approach on the standard Fashion-MINST dataset, comparing it to various regularization techniques. Interestingly, we note that by nudging each neuron to divide, at least in part, its input data, the resulting networks make use of each neuron, avoiding a hyperplane completely on one side of its input data (which is equivalent to a constant into the next layers). To illustrate this point, we study the prevalence of saturated nodes throughout training, showing that neurons are activated more frequently and earlier in training when using this regularization approach. A direct consequence of the improved neuron activation is that deep networks are now easier to train. This is crucially important when the network topology is not known a priori and fitting often remains stuck in a suboptimal local minima. We demonstrate this property by training a network of increasing depth (and constant width): most regularization approaches will result in increasingly frequent training failures (over different random seeds) whilst the proposed evidence-based regularization significantly outperforms in its ability to train deep networks.

In many Bayesian computations, first, we obtain the expression of the joint distribution of all the unknown variables given the observed data. In general, this expression is not separable in those variables. Thus, obtaining their marginals for each variable and computing the expectations are difficult and costly. This problem becomes even more difficult in high dimensional quandaries, which is an important issue in inverse problems. We may then try to propose a surrogate expression with which we can do approximate computations. Often a separable expression approximation can be useful enough. The Variational Bayesian Approximation (VBA) is a technique that approximates the joint distribution $p$ with an easier, for example separable, one $q$ by minimizing Kullback–Leibler Divergence $KL(q|p)$. When $q$ is separable in all the variables, the approximation is also called Mean Field Approximation (MFA) and so $q$ is the product of the approximated marginals. A first standard and general algorithm is alternate optimization of $KL(q|p)$ with respect to $q_i$. A second general approach is its optimization in the Riemannian manifold. However, in this paper, for practical reasons, we consider the case where $p$ is in the exponential family and so is $q$. For this case, $KL(q|p)$ becomes a function of the parameters $\thetab$ of the exponential family. Then, we can use any other optimization algorithm to obtain those parameters. In this paper, we compare three optimization algorithms: standard alternate optimization, a gradient-based algorithm and a natural gradient algorithm and study their relative performances on three examples.

Measurement uncertainty is typically expressed in terms of a symmetric interval , where denotes the measurement result and the expanded uncertainty. However, in the case of heteroscedasticity, symmetric uncertainty intervals can be misleading. In this paper, a different approach for the calculation of uncertainty intervals is introduced. This approach is applicable when a validation study has been conducted with samples with known concentrations. It will be shown how, under certain circumstances, asymmetric uncertainty intervals arise quite naturally and lead to more reliable uncertainty intervals.

This paper focuses on the so-called proportional intensity-based software reliability models (PI-SRMs), which are extensions of the common non homogeneous Poisson process (NHPP)-based SRMs, and describe the probabilistic behavior of software fault-detection process by incorporating the time-dependent software metrics data observed in the development process. Especially we generalize the seminal PI-SRM in Rinsaka, Shibata and Dohi (2006) by introducing eleven well-known fault-detection time distributions, and investigate their goodness-of-fit and predictive performances. In numerical illustrations with four data sets collected in real software development projects, we utilize the maximum likelihood estimation to estimate model parameters with three time-dependent covariates; test execution time, failure identification work and computer time-failure identification, and examine the performances of our PI SRMs in comparison with the existing NHPP-based SRMs without covariates. It is shown that our PI-STMs could give better goodness-of-fit and predictive performances in many cases.

Developing new ways to estimate probabilities can be valuable for science, statistics, and engineering. By considering the information content of different output patterns, recent work invoking algorithmic information theory has shown that a priori probability predictions based on pattern complexities can be made in a broad class of input-output maps. These algorithmic probability predictions do not depend on a detailed knowledge of how output patterns were produced, or historical statistical data. Although quantitatively fairly accurate, a main weakness of these predictions is that they are given as an upper bound on the probability of a pattern, but many low complexity, low probability patterns occur, for which the upper bound has little predictive value. Here we study this low complexity, low probability phenomenon by looking at example maps, namely a finite state transducer, natural time series data, RNA molecule structures, and polynomial curves. Some mechanisms causing low complexity, low probability behaviour are identified, and we argue this behaviour should be assumed as a default in the real world algorithmic probability studies. Additionally, we examine some applications of algorithmic probability and discuss some implications of low complexity, low probability patterns for several research areas including simplicity in physics and biology, a priori probability predictions, Solomonoff induction and Occam's razor, machine learning, and password guessing.

The twisted Wang transform distribution family, defined as the composition of parameter shifted inverse CDF function with an original CDF function, is found to be most suitable for matching low shape factor distributions, characterizing hard to fit or to simulate reinsurance portfolio losses for some perils from our previous study. Among them, the best form for matching a hard-to-fit empirical loss distribution for a specific peril, is the Exponential Fractional Extra Power 0 Distribution in (0,1) with CDF:.The simplest yet still a good form of this family is the Transformed Hyperbolic Tangent Distribution with CDF:,which has analytical formulas for the moments. The twisted Wang transform distribution family is compared and confirmed to be superior to all other well-known distribution families through extensive numerical optimization practice, distribution forms guesses, and computer-aided exploration experiments.

For the symmetrical mixture model and mixture test area, the lattice point set is used to partition, and then the corresponding test statistics can be constructed. In this paper, we first proposes the partition methods under the lattice point sets and obtains several sub-simplexes without common interior points. Furthermore, we present the method for constructing a uniform design on the simplex using the center points of these sub-simplexes. The designs satisfy the uniformity of maximum distance deviation and provide good results for the mean square error deviation. Finally, the uniformity test on the mixture region is considered and illustrated by examples.

Two extensions of the generalized double Lomax distribution are introduced - the spliced-scale generalized double Lomax distribution and the exponentiated generalized double Lomax distribution. Their properties are studied. Usefulness of the new distributions is shown by fitting them to stock returns datasets.

Considering that a fair coin has two sides and a cylindrical edge, the probability that it would fall on its edge is calculated, yielding the probability of heads or tails less than 50%. Theoretical models for a static case and for five dynamic cases, without and with rebounds, show that there is a small probability that the coin does not fall on its head or tail, depending on initial toss conditions, the coin geometry and conditions of the coin and landing surfaces. For the dynamic model with rebounds, it is found that the probability that a 50 Eurocent coin thrown from a normal height with common initial velocity conditions and appropriate surface conditions will end up on its edge is in the order of one against several thousand.

In this note, we provide some effective treatments of a general linear model with adding-up restrictions via algebraic operations of given vectors and matrices in the model, including analytic expressions of the well-known ordinary least-squares estimator (OLSE) and the best linear unbiased estimator (BLUE) of the unknown parameters in the model.

As its capital, Jakarta plays a critical role in boosting Indonesia’s economic growth and setting the precedent for broader change outside of the city. One crucial avenue of inquiry to better understand, and prepare for, the future of a country so heavily impacted by disastrous weather events is understanding the effects of climate change through data. This study investigates meteorological data collected from 1996 to 2021 and compares the application of the SARIMA and the Holt-Winters methods to predict the future influence of climatic parameters on Jakarta’s weather. The performance of the SARIMA method is proven to provide better results than the Holt-Winter models and both methods showed the best performances when forecasting the humidity data. The results of the forecast are able to demonstrate the characteristic of the climate in Jakarta, with dry season ranging from May to October and wet season ranging from November to April.

Automated software bots infiltrate online surveys and corrupt data integrity, not to mention waste researcher time and budgets. Although resources exist to help keep bots out and identify bots when they do evade survey barriers, bot attacks may be a persistent problem for online surveys for a long time to come. Bots are evolving -- even as survey designers try ever more sophisticated methods to fend them off and weed their answers out. Vigilance needs to be high and the bot generators should not be under-estimated. We recount here some bot features we encountered after our own survey was attacked that helped to identify them, and that have not been detailed elsewhere. We also discuss reasons why commonly recommended strategies for how to keep bots out may not be feasible for many scientific researchers.

Objectives: This paper presents a statistical review of modelling simulations for frequency and sensitivity of COVID-19 testing paradigms. Methods: We performed a review of preprints and published articles on PubMed from January 1, 2020 – March 1, 2021 using the search terms “COVID screening testing”, “COVID testing frequency”, “COVID testing frequency screening” and “SARS-CoV-2 testing frequency”.Results: Several authors’ conclusions support the claim that test frequency and test sensitivity both play a role in reducing SARS-CoV-2 transmission. We highlight the interplay between frequency of testing, test sensitivity and the speed at which test results are available in our review. Conclusions: Evidence suggests that sensitivity and frequency of testing both play a part in decreasing transmission of disease. We conclude that, overall, test sensitivity plays less of a role in reducing disease transmission in a population compared to the frequency of testing and how quickly test results are available.

A paradigm shift away from null hypothesis significance testing seems in progress. Based on simulations, we illustrate some of the underlying motivations. First, P-values vary strongly from study to study, hence dichotomous inference using significance thresholds is usually unjustified. Second, statistically significant results have overestimated effect sizes, a bias declining with increasing statistical power. Third, statistically non-significant results have underestimated effect sizes, and this bias gets stronger with higher statistical power. Fourth, the tested statistical hypotheses usually lack biological justification and are often uninformative. Despite these problems, a screen of 48 papers from the 2020 volume of the Journal of Evolutionary Biology exemplifies that significance testing is still used almost universally in evolutionary biology. All screened studies tested default null hypotheses of zero effect with the default significance threshold of p = 0.05, none presented a pre-specified alternative hypothesis, pre-study power calculation, and the probability of ‘false negatives’(beta error). The results sections of the papers presented 49 significance tests on average (median 23, range 0–390). Of 41 studies that contained verbal descriptions of a ‘statistically non-significant’ result, 26 (63%) falsely claimed the absence of an effect. We conclude that studies in ecology and evolutionary biology are mostly exploratory and descriptive. We should thus shift from claiming to ‘test’ specific hypotheses statistically to describing and discussing many hypotheses (possible true effect sizes) that are most compatible with our data, given our statistical model. We already have the means for doing so, because we routinely present compatibility (‘confidence’) intervals covering these hypotheses.

This paper explores the tuning and results of two-part models on rich datasets provided through the Casualty Actuarial Society (CAS). These data sets include BI (bodily injury), PD (property damage) and COLL (collision) coverage, each documenting policy characteristics and claims across a four year period. The datasets are explored, including summaries of all variables, then the methods for modeling are set forth. Models are tuned and the tuning results are displayed, after which we train the final models and seek to explain select predictions. All of the code will be made available on GitHub. Data was provided by a private insurance carrier to the CAS after anonymizing the data set. This data is available to actuarial researchers for well-defined research projects that have universal benefit to the insurance industry and the public. Our hope is that the methods demonstrated here can be a good foundation for future ratemaking models to be developed and tested more efficiently.

Stage of cancer is a discrete ordinal response that indicates aggressiveness of disease and is often used by physicians to determine the type and intensity of treatment to be administered. For example, the FIGO stage in cervical cancer is based on the size and depth of the tumor as well as the level of spread. It may be of clinical relevance to identify molecular features from high-throughput genomic assays that are associated with stage of cervical cancer, to elucidate pathways related to tumor aggressiveness, identify improved molecular features that may be useful for staging, and identify therapeutic targets. High-throughput RNA-Seq data and corresponding clinical data (including stage) for cervical cancer patients has been made available through The Cancer Genome Atlas Project (TCGA). We recently described penalized Bayesian ordinal response models that can be used for variable selection for over-parameterized datasets such as the TCGA-CESC dataset. Herein, we describe our ordinalbayes R package, available from the Comprehensive R Archive Network (CRAN), which is capable of fitting cumulative logit models when the outcome is ordinal and the number of predictors exceeds the sample size, P>N, such as for TCGA data. We demonstrate use of this package through application to TCGA cervical cancer dataset. Our ordinalbayes package can be used to fit models to high-dimensional dataset and effectively performs variable selection.

Regression models are mostly used in all fields of sciences for modelling the relationship between a dependent variable and independent variable(s). The least square method is often used to estimate the parameters in a linear model because it is the best linear unbiased estimator. These estimates can only be reliable if the assumption of normality is satisfied. In some cases the dependent variable might be bimodal and shows a non-linear relationship with the independent variable(s). In this case, a non-linear model should be used. In non-linear model, the standard errors are often obtained by linearizing the nonlinear function around the parameter, assuming central limit theorem. After the linearization, the least square parameter estimates are obtained. It should be noted that the error of the non-linear model is different from that of the transformed linear model. Thus, there is a need to transform back to the original non-linear model. In this note, a novel non-linear function was developed into a non-linear regression model, called Normal-Power model. The least square method was used to estimate the parameter of the transformed model. Its usefulness in regression model was demonstrated using real data of Nigeria Economy-Tourism model.

In this work, we introduce a generalized measure of cumulative residual entropy and study its properties. We show that several existing measures of entropy such as cumulative residual entropy, weighted cumulative residual entropy and cumulative residual Tsallis entropy, are all special cases of the generalized cumulative residual entropy. We also propose a measure of generalized cumulative entropy, which includes cumulative entropy, weighted cumulative entropy and cumulative Tsallis entropy as special cases. We discuss generating function approach using which we derive different entropy measures. Finally, using the newly introduced entropy measures, we establish some relationships between entropy and extropy measures.

Fitting parameters on spline curves produces more parsimonious models, maintaining fit quality. Smoothing the splines reduces predictive variance. Individual splines are often fit by type of variable, e.g., in age-period-cohort models. Linear and cubic splines are most common. Several smoothing criteria have been used for parameter curves, with cubic splines fit by constraining the integral of splines’ squared-second derivatives popular recently. Constraining the sum of squared second differences for linear splines is analogous. Generally the degree of smoothing is selected using cross-validation. Known spline dummy-variable matrices allow regression estimation of splines, with smoothing done via constrained regression. Smoothing criteria based on sums of squares or absolute values of parameters, as in ridge regression or LASSO, improves predictive accuracy and produces splines similar to smoothing by second-derivative constraints. Variables with very low t-statistics represent points where curve-shapes barely change. Eliminating those variables leaves knots concentrated where spline shapes change. A Bayesian version of this puts shrinkage priors on spline parameters. This yields realistic joint parameter distributions, avoids problems associated with using cross-validation for parameter estimation, and readily expands to non-linear modeling, such as interactions among variable types. Regularized regression and Bayesian spline methods are compared for two example datasets.

The present era of the Internet of Things (IoT) having intelligent functionalities in solving problems pertaining to realtime mission-critical systems has brought an immense revolution in diverse fields including healthcare and navigation systems. However, to the best of our knowledge, the potential of IoT has not been fully exploited yet in the field of the energy sector. We argue that there is an immense need to shift the traditional mission-critical electric power system architecture to IoT-based fully orchestrated architecture in order to increase efficiency, as billions of investment is reserved for the energy sector globally. Since network orchestration deals with auomating the interaction between multiple components involved to execute a particular service, therefore, scheduling the relevant processes within strict deadlines becomes the core pillar of the architecture. The mission-critical systems with urgent task execution often suffer from issues of missing task deadlines. In this study, we present a novel IoT task orchestration architecture for efficient energy management of a nanogrid system that focuses on minimizing the use of nonrenewable energy resources and maximizing the use of renewable energy resources. Moreover, major components of IoT task orchestration such as task mapping and task scheduling are also enhanced using NLP and PSO optimization modules. The proposed task scheduling algorithm incorporates the optimized surplus time, and efficiently executes the energy management-related tasks contemplating to their types. The study utilizes sensors to obtain data from physical IoT devices, including photovoltaic (PV), Energy Storage System (ESS), and diesel generator (DG). The performance of the proposed model is evaluated using data set of nanogrid houses. The outcomes revealed that IoT-task orchestration has played a pivotal role in efficient energy management for nanogrid mission-critical system. Furthermore, the comparison with state-of-the-art scheduling algorithms showed that the task starvation rate is reduced to 16% and 12% when compared with RR and FEF algorithms, respectively.

To detect aberrant human behaviors from large volume of passive data collected by smartphones in real time, we propose an online anomaly detection method using Hotelling’s T-squared test. The test statistic is a weighted average, with more weight on the between-individual component when there are little data available for the individual and more weight on the within-individual component when the data are adequate. The algorithm takes only O(1) run time in each update and the required memory usage is fixed after a pre-specified number of updates. The performance of the proposed method, in terms of accuracy, sensitivity and specificity, is consistently better than or equal to the offline method that it builds upon depending on the sample size of the individual data.

Bayesian nonlinear mixed effects models for data in the form of continuous, repeated measurements from a population, also known as Bayesian hierarchical nonlinear models, are a popular platform for analysis when interest focuses on individual specific characteristics and relevant uncertainty quantification. Due to the limitation of computational power, this framework was relatively dormant until the late 1980s, but in recent years, the statistical research community saw vigorous development of new methodological and computational techniques for these models, the emergence of software, and wide application of the models in numerous industrial and academic fields. This article presents an overview of the formulation, interpretation, and implementation of Bayesian nonlinear mixed effects models and surveys recent advances and applications.

The purpose of this study was to identify the factors that affect the mortality among adult HIV/TB co-infected patients and to see the nutritional difference among mortality in residence level. Retrospective cohort studies of 417 patients which fulfill our criteria were included. Multilevel logistic regression models were used. MLwiN and SPSS software are used to estimate the parameter. The variance of the random factor in the empty model was significant which indicates that there were residential differences in TB-HIV co-infected mortality and it shows multilevel analysis was an appropriate approach for further analysis. The prevalence of HIV/TB co-infected patients' death was 12.9% in study time. Functional status, age of patients, WHO clinical stages, nutritional status, CD4 counts, regimen, and BMI were found to be significant determinants of HIV/TB co-infected mortality. In our study, patients with the bedridden category of functional status, the fourth stages of WHO clinical stages (stage IV), patients with higher age, patients whose treatments were second-line regimen and low CD4 cell counts were more at risk of death. The study also revealed that; poor nutritional status increased the risk of mortality among HIV/TB co-infected patients and it varies among the residence of the patients (rural area were more at risk).

3D facial surface imaging is a useful tool in dentistry and in terms of diagnostics and treatment planning. Between-groups PCA (bgPCA) is a method that has been used to analyse shapes in biological morphometrics, although various “pathologies” of bgPCA have recently been proposed. Monte Carlo (MC) simulated datasets were created here in order to explore “pathologies” of multilevel PCA (mPCA), where mPCA with two levels is equivalent to bgPCA. The first set of MC experiments involved 300 uncorrelated normally distributed variables, whereas the second set of MC experiments used correlated multivariate MC data describing 3D facial shape. We confirmed previous results of other researchers that indicated that bgPCA (and so also mPCA) can give a false impression of strong differences in component scores between groups when there is none in reality. These spurious differences in component scores via mPCA reduced strongly as the sample sizes per group were increased. Eigenvalues via mPCA were also found to be strongly effected by imbalances in sample sizes per group, although this problem was removed by using weighted forms of covariance matrices suggested by the maximum likelihood solution of the two-level model. However, this did not solve problems of spurious differences between groups in these simulations, which was driven by very small sample sizes in one group here. As a “rule of thumb” only, all of our experiments indicate that reasonable results are obtained when sample sizes per group in all groups are at least equal to the number of variables. Interestingly, the sum of all eigenvalues over both levels via mPCA scaled approximately linearly with the inverse of the sample size per group in all experiments. Finally, between-group variation was added explicitly to the MC data generation model in two experiments considered here. Results for the sum of all eigenvalues via mPCA predicted the asymptotic amount for the total amount of variance correctly in this case, whereas standard “single-level” PCA underestimated this quantity.

Classification predictive modeling involves the accurate assignment of observations in a dataset to target classes or categories. There is an increasing growth of real-world classification problems with severely imbalanced class distributions. In this case, minority classes have much fewer observations to learn from than those from majority classes. Despite this sparsity, a minority class is often considered the more interesting class yet developing a scientific learning algorithm suitable for the observations presents countless challenges. In this article, we suggest a novel multi-class classification algorithm specialized to handle severely imbalanced classes based on the method we refer to as SAMME.C2. It blends the flexible mechanics of the boosting techniques from SAMME algorithm, a multi-class classifier, and Ada.C2 algorithm, a cost-sensitive binary classifier designed to address highly class imbalances. Not only do we provide the resulting algorithm but we also establish scientific and statistical formulation of our proposed SAMME.C2 algorithm. Through numerical experiments examining various degrees of classifier difficulty, we demonstrate consistent superior performance of our proposed model.

Let $f(t)$ be a characteristic function. The question on infinite divisibility of $g_{2k}(t)=f^{(2k)}(t)/f^{(2k)}(0)$ is considered. There are given the condition for that function not to be infinite divisible. Some examples of infinite divisibility of $g_{2k}(t)$ are given.

A single Information security is of pivotal concern for consistently streaming information over the widespread internetwork. The bottleneck flow of incoming and outgoing data traffic introduces the issue of malicious activities taken place by intruders, hackers and attackers in the form of authenticity desecration, gridlocking data traffic, vandalizing data and crashing the established network. The issue of emerging suspicious activities is managed by the domain of Intrusion Detection Systems (IDS). The IDS consistently monitors the network for identifica-tion of suspicious activities and generates alarm and indication in presence of malicious threats and worms. The performance of IDS is improved by using different signature based machine learning algorithms. In this paper, the performance of IDS model is determined using hybridization of nestrov-accelerated adaptive moment estimation –stochastic gradient descent (HNADAM-SDG) algorithm. The performance of the algorithm is compared with other classi-fication algorithms as logistic regression, ridge classifier and ensemble algorithm by adapting feature selection and optimization techniques

This paper considers the solution of the equations for ruin probabilities in infinite continuous time. Using the Fourier Transform and certain results from the theory of complex functions, these solutions are obtained as com- plex integrals in a form which may be evaluated numerically by means of the inverse Fourier Transform. In addition the relationship between the re- sults obtained for the continuous time cases, and those in the literature, are compared. Closed form ruin probabilities for the heavy tailed distributions: mixed exponential; Gamma (including Erlang); Lognormal; Weillbull; and Pareto, are derived as a result (or computed to any degree of accuracy, and without the use of simulations).

The relative operating characteristic (ROC) curve is a popular diagnostic tool in forecast verification, with the area under the ROC curve (AUC) used as a verification metric measuring the discrimination ability of a forecast. Along with calibration, discrimination is deemed as a fundamental probabilistic forecast attribute. In particular, in ensemble forecast verification, AUC provides a basis for the comparison of potential predictive skill of competing forecasts. While this approach is straightforward when dealing with forecasts of common events (e.g. probability of precipitation), the AUC interpretation can turn out to be oversimplistic or misleading when focusing on rare events (e.g. precipitation exceeding some warning criterion). How should we interpret AUC of ensemble forecasts when focusing on rare events? How can changes in the way probability forecasts are derived from the ensemble forecast affect AUC results? How can we detect a genuine improvement in terms of predictive skill? Based on verification experiments, a critical eye is cast on the AUC interpretation to answer these questions. As well as the traditional trapezoidal approximation and the well-known bi-normal fitting model, we discuss a new approach which embraces the concept of imprecise probabilities and relies on the subdivision of the lowest ensemble probability category.

Background: In the functional data analysis (FDA), the hybrid or mixed data are scalar and functional datasets. The semi-functional partial linear regression model (SFPLR) is one of the first semiparametric models for the scalar response with hybrid covariates. Various extensions of this model are explored and summarized. Methods: Two first research articles, including “semi-functional partial linear regression model”, and “Partial functional linear regression” have more than 300 citations in Google Scholar. Finally, only 106 articles remained according to the inclusion and exclusion criteria such as 1) including the published articles in the ISI journals and excluding 2) non-English and 3) preprints, slides, and conference papers. We use the PRISMA standard for systematic review. Results: The articles are categorized into the following main topics: estimation procedures, confidence regions, time series, and panel data, Bayesian, spatial, robust, testing, quantile regression, varying Coefficient Models, Variable Selection, Single-index model, Measurement error, Multiple Functions, Missing values, Rank Method and Others. There are different applications and datasets such as the Tecator dataset, air quality, electricity consumption, and Neuroimaging, among others. Conclusions: SFPLR is one of the most famous regression modeling methods for hybrid data that has a lot of extensions among other models.

We address the process of discounting in random environments which allows to value the far future in economic terms. We review several approaches to the problem regarding different well-established stochastic market dynamics in the continuous-time context and include the Feynman-Kac approach. We also review the relation between bond pricing theory and discount and introduce the market price of risk and the risk neutral measures from an intuitive point of view devoid of excessive formalism. We provide the discount for each economic model and discuss their key results. We finally present a summary of our previous empirical studies on several countries of the long-run discount problem.

Decision tree algorithms have been among the most popular algorithms for interpretable (transparent) machine learning since the early 1980s. On the other hand, deep learning methods have boosted the capacity of machine learning algorithms and are now being used for non-trivial applications in various applied domains. But training a fully-connected deep feed-forward network by gradient-descent backpropagation is slow and requires arbitrary choices regarding the number of hidden units and layers. In this paper, we propose near-optimal neural regression trees, intending to make it much faster than deep feed-forward networks and for which it is not essential to specify the number of hidden units in the hidden layers of the neural network in advance. The key idea is to construct a decision tree and then simulate the decision tree with a neural network. This work aims to build a mathematical formulation of neural trees and gain the complementary benefits of both sparse optimal decision trees and neural trees. We propose near-optimal sparse neural trees (NSNT) that is shown to be asymptotically consistent and robust in nature. Additionally, the proposed NSNT model obtain a fast rate of convergence which is near-optimal up to some logarithmic factor. We comprehensively benchmark the proposed method on a sample of 80 datasets (40 classification datasets and 40 regression datasets) from the UCI machine learning repository. We establish that the proposed method is likely to outperform the current state-of-the-art methods (random forest, XGBoost, optimal classification tree, and near-optimal nonlinear trees) for the majority of the datasets.

Three-dimensional reconstruction plays an important role in assisting doctors and surgeons in diagnosing bone defects’ healing progress. Common three-dimensional reconstruction methods include surface and volume rendering. As the focus is on the shape of the bone, volume rendering is omitted. Many improvements have been made on surface rendering methods like Marching Cubes and Marching Tetrahedra, but not many on working towards real-time or near real-time surface rendering for large medical images, and studying the effects of different parameter settings for the improvements. Hence, in this study, an attempt towards near real-time surface rendering for large medical images is made. Different parameter values are experimented on to study their effect on reconstruction accuracy, reconstruction and rendering time, and the number of vertices and faces. The proposed improvement involving three-dimensional data smoothing with convolution kernel Gaussian size 0.5 and mesh simplification reduction factor of 0.1, is the best parameter value combination for achieving a good balance between high reconstruction accuracy, low total execution time, and a low number of vertices and faces. It has successfully increased the reconstruction accuracy by 0.0235%, decreased the total execution time by 69.81%, and decreased the number of vertices and faces by 86.57% and 86.61% respectively.

We investigate the power of time series analysis based on a variety of information-theoretic approaches from statistics (AIC, BIC) and machine learning (Minimum Message Length) - and we then compare their efficacy with traditional time series model and with hybrids involving deep learning. More specifically, we develop AIC, BIC and Minimum Message Length (MML) ARMA (autoregressive moving average) time series models - with this Bayesian information-theoretic MML ARMA modelling already being new work. We then study deep learning based algorithms in time series forecasting, using Long Short Term Memory (LSTM), and we then combine this with the ARMA modelling to produce a hybrid ARMA-LSTM prediction. Part of the purpose of the use of LSTM is to seek capture any hidden information in the residuals left from the traditional ARMA model. We show that MML not only outperforms earlier statistical approaches to ARMA modelling, but we further show that the hybrid MML ARMA-LSTM models outperform both ARMA models and LSTM models.

We propose a new ensemble classification algorithm, named Super Random Subspace Ensemble (Super RaSE), to tackle the sparse classification problem. The proposed algorithm is motivated by the Random Subspace Ensemble algorithm (RaSE). The RaSE method was shown to be a flexible framework that can be coupled with any existing base classification. However, the success of RaSE largely depends on the proper choice of the base classifier, which is unfortunately unknown to us. In this work, we show that Super RaSE avoids the need to choose a base classifier by randomly sampling a collection of classifiers together with the subspace. As a result, Super RaSE is more flexible and robust than RaSE. In addition to the vanilla Super RaSE, we also develop the iterative Super RaSE, which adaptively changes the base classifier distribution as well as the subspace distribution. We show the Super RaSE algorithm and its iterative version perform competitively for a wide range of simulated datasets and two real data examples. The new Super RaSE algorithm and its iterative version are implemented in a new version of the R package RaSEn.

In many practical applications, it is more convenient to characterize the quality of production processes or service operations throughout the count of nonconformities. In the context of SPC, nonconformities are usually assumed to appear according to the binomial probability model. The conventional way for monitoring nonconformities involves Shewhart-type control procedures based on both constant and time-varying sample sizes. In this article, an EWMA scheme is proposed for monitoring the fraction of nonconforming items with time-varying sample sizes. The proposed control chart is referred to as the EWMAG-B and can be easily adapted to work with a constant sample size by fixing it at a needed value. By means of simulation, it was found out that the EWMAG-B chart outperforms the conventional p control chart in Phase II while detecting changes in the process level is wanted.

This work addresses the problem of sampling from Gaussian probability distributions by means of uniform samples obtained deterministically and directly from space-filling curves (SFCs), a purely topological concept. To that end, the well-known inverse cumulative distribution function method is used, with the help of the probit function,which is the inverse of the cumulative distribution function of the standard normal distribution. Mainly due to the central limit theorem, the Gaussian distribution plays a fundamental role in probability theory and related areas, and that is why it has been chosen to be studied in the present paper. Numerical distributions (histograms) obtained with the proposed method, and in several levels of granularity, are compared to the theoretical normal PDF, along with other already established sampling methods, all using the cited probit function. Final results are validated with the Kullback-Leibler and two other divergence measures, and it will be possible to draw conclusions about the adequacy of the presented paradigm. As is amply known, the generation of uniform random numbers is a deterministic simulation of randomness using numerical operations. That said, sequences resulting from this kind of procedure are not truly random. Even so, and to be coherent with the literature, the expression ”random number” will be used along the text to mean ”pseudo-random number”.

A major task in Functional Time Series Analysis is measuring the dependence within and between processes, for which lagged covariance and cross-covariance operators have proven to be a practical tool in well-established spaces. This article deduces estimators and asymptotic upper bounds of the estimation errors for lagged covariance and cross-covariance operators of processes in Cartesian products of abstract Hilbert spaces for fixed and increasing lag and Cartesian powers. We allow the processes to be non-centered, and to have values in different spaces when investigating the dependence between processes. Also, we discuss features of estimators for the principle components of our covariance operators.

Evidence-based medicine (EBM) is in crisis, in part due to bad methods, which are understood as misuse of statistics that is considered correct in itself. This article exposes two related common misconceptions in statistics, the effect size (ES) based on correlation (CBES) and a misconception of contingency tables (MCT). CBES is a fallacy based on misunderstanding of correlation and ES and confusion with 2 × 2 tables, which makes no distinction between gross crosstabs (GCTs) and contingency tables (CTs). This leads to misapplication of Pearson’s Phi, designed for CTs, to GCTs and confusion of the resulting gross Pearson Phi, or mean-square effect half-size, with the implied Pearson mean square contingency coefficient. Generalizing this binary fallacy to continuous data and the correlation in general (Pearson’s r) resulted in flawed equations directly expressing ES in terms of the correlation coefficient, which is impossible without including covariance, so these equations and the whole CBES concept are fundamentally wrong. MCT is a series of related misconceptions due to confusion with 2 × 2 tables and misapplication of related statistics. The misconceptions are threatening because most of the findings from contingency tables, including CBES-based meta-analyses, can be misleading. Problems arising from these fallacies are discussed and the necessary changes to the corpus of statistics are proposed resolving the problem of correlation and ES in paired binary data. Since exposing these fallacies casts doubt on the reliability of the statistical foundations of EBM in general, we urgently need to revise them.

In statistics and probability theory, one of the most important statistics is the sums of random variables. After introducing a probability distribution, determining the sum of n independent and identically distributed random variables is one of the interesting topics for authors. This paper presents the probability density functions for the sum of n independent and identically distributed random variables such as Shanker, Akash, Ishita, Pranav, Rani, and Ram Awadh. In order to determine all aforementioned distributions, the problem-solving methods are applied which is based on the change-of-variables technique.

We are of the opinion that during the design of a binary classifier one ought to consider adding an “I don’t know” answer. We provide the case for the introduction of this third category when a human needs to make a decision based on the answer from a binary classifier. We discuss the costs and potential benefits of its introduction. Colloquially, we have used the term “I don’t know”, but formally we refer to it as NotAvailable. A procedure to define NotAvailable predictions in binary classifiers, called all leave-one-out models (ALOOM), is presented as proof of the concept. Furthermore, we discuss the potential benefits of applying ALOOM in real life applications.

While entropy was introduced in the second half of the 19th century in the international vocabulary as a scientific term, in the 20th century it became common in colloquial use. Popular imagination has loaded “entropy” with almost every negative quality in the universe, in life and in society, with a dominant meaning of disorder and disorganization. Exploring the history of the term and many different approaches on it, we show that entropy has a universal stochastic definition which is not disorder. The accompanying principle of maximum entropy, which lies behind the Second Law, gives explanatory and inferential power to the concept and promotes entropy as the mother of creativity and evolution. As social sciences are often contaminated by subjectivity and ideological influences, we try to explore whether the maximum entropy, applied to the distribution of wealth quantified by annual income, can give an objective description. Using publicly available income data, we show that the income distribution is consistent with the principle of the maximum entropy. The increase of entropy is associated to increase of society’s wealth yet a standardized form of entropy can be used to quantify inequality. Historically, technology has played a major role in development and increase of the entropy of income. Such findings are contrary to the theories of ecological economics and other theories which use the term entropy in a Malthusian perspective.

A class of models for non-Gaussian spatial random fields is explored for spatial field reconstruction in environmental and sensor network monitoring. The family of models explored utilises a class of transformation functions known as the Tukey g-and-h transformations to create a family of warped spatial Gaussian process models which can support various desirable features such as flexible marginal distributions, which can be skewed, leptokurtic and/or heavy-tailed. The resulting model is widely applicable in a range of spatial field reconstruction applications. To utilise the model in applications in practice, it is important to carefully characterise the statistical properties of the Tukey g-and-h random fields. In this work, we both study the properties of the resulting warped Gaussian processes as well as using the characterising statistical properties of the warped processes to obtain flexible spatial field reconstructions. In this regard, we derive five different estimators for various important quantities often considered in spatial field reconstruction problems. These include the multi-point Minimum Mean Squared Error (MMSE) estimators; the multiple point Maximum A-Posteriori (MAP) estimators; an efficient class of multiple-point linear estimators based on the Spatial-Best Linear Unbiased (S-BLUE) estimators; and two multi-point threshold exceedance based estimators, namely the Spatial Regional and Level Exceedance estimators. Simulation results and real data examples show the benefits of using the Tukey g-and-h transformation as opposed to standard Gaussian spatial random fields in a real data application for environmental monitoring.

The power function distribution is a flexible waiting time model that may provide better fit for some failure data. This paper presents the comparison of the maximum likelihood estimates and the Bayes estimates of two-parameter power function distribution. The Bayes estimates are obtained, using conjugate priors, under five loss functions consist of square error, precautionary, weighted, LINEX and DeGroot loss function. The Gibbs sampling algorithm is proposed to generate samples from posterior distributions and in result the Bayes estimates are computed. The comparison of the maximum likelihood estimates and the Bayes estimates are done through the root mean squared errors. One real-life data set is analyzed to illustrate the evaluation of proposed methods of estimation. Finally, results from the simulation are discussed to assess the performance behavior of the maximum likelihood estimates and the Bayes estimates.

A method for noise reduction of spectra based on the adaptive application of the Savitsky-Golay polynomial filter is presented. A polynomial approximation is calculated at all points of the spectrum and for all window sizes. The weighted sum of all polynomials containing the point to be processed is used as the result. The weighting factors are calculated by evaluating the quality of the fit. This paper proposes two evaluation functions. The performance of the presented method is compared with the Savitsky-Golay method and the wavelet noise reduction method. The proposed approach provides good noise reduction performance without using user-entered parameters.

Influential Observations, which are outliers in x direction, y direction or both, remain a hitch in classical regression model fitting. Spatial regression model, with peculiar nature of outliers due to their local nature, is not free from the effect of such influential observations. Researchers have adapted some classical regression techniques to the spatial models and yielded satisfactory results. However, masking or/and swamping remain stumbling block to such methods. We obtained the spatial representation of the classical regression measures of diagnostic in general spatial model. Commonly used diagnostic measure in spatial diagnostic, the Cook's distance, is compared to some robust methods, Hi2 (using robust and non-robust measures), and classification based on generalized residuals and diagnostic generalized potentials, ISRs-Posi and ESRs-Posi, with the help of the obtained spatial prediction residuals and the spatial leverage term. Results of simulation and applications to real data have shown the advantage of the ISRs-Posi and ESRs-Posi due to classification of outliers over Cook's distance and non-robust Hsi12, which suffer from masking, and robust Hsi22 which suffer from swamping in general spatial model.

Using Directed Transfer Function (DTF) and Partial Directed Coherence (PDC) in their information version, this paper extends their theoretical framework to incorporate instantaneous Granger Causality (iGC)’s frequency domain description into a single unified perspective. We show that standard vector autoregressive models allow portraying iGC’s repercussions associated with Granger Connectivity where interactions mediated without delay between time series can be easily detected.

In this research first, a sequence of properties called delta is assigned to each prime number and then examined. Deltas are only dependent on the distribution of prime numbers, so the results obtained for the delta distribution can be considered as a proxy for the distribution of prime numbers. The first observation was that these properties are not unique and different prime numbers may have the same value of delta of a given order. It was found that a small number of deltas cover a large portion of prime numbers, so by recognizing repetitive deltas, the next prime numbers can be predicted with a certain probability, but the most important observation of this study is the normal distribution of deltas. This research has not tried to justify the obtained observations and instead of answering the questions, it seeks to ask the right question.

The goal of clustering is to identify common structures in a data set by forming groups of homogeneous objects. The observed characteristics of many economic time series motivated the development of classes of distributions that can accommodate properties such as heavy tails and skewness. Thanks to its flexibility, the Skewed Exponential Power Distribution (also called Skewed Generalized Error Distribution) ensures a unified and general framework for clustering possibly skewed and heavy tailed time series. This paper develops a clustering procedure of model-based type, assuming that the time series are generated by the same underlying probability distribution but with different parameters. Moreover, we propose to optimally combine the estimated parameters to form the clusters with an entropy weighing $k$-means approach. The usefulness of the proposal is showed by means of application to financial time series, demonstrating also how the obtained clusters can be used to form portfolio of stocks.

We outline and test a new methodology for genuine simulation of stochastic processes with any dependence and any marginal distribution. We reproduce time dependence with a generalized, time symmetric or asymmetric, moving-average scheme. This implements linear filtering of non-Gaussian white noise, with the weights of the filter determined by analytical equations in terms of the autocovariance of the process. We approximate the marginal distribution of the process, irrespective of its type, using a number of its cumulants, which in turn determine the cumulants of white noise in a manner that can readily support the generation of random numbers from that approximation, so that it be applicable for stochastic simulation. The simulation method is genuine as it uses the process of interest directly without any transformation (e.g. normalization). We illustrate the method in a number of synthetic and real-world applications with either persistence or antipersistence, and with non-Gaussian marginal distributions that are bounded, thus making the problem more demanding. These include distributions bounded from both sides, such as uniform, and bounded form below, such as exponential and Pareto, possibly having a discontinuity at the origin (intermittence). All examples studied show the satisfactory performance of the method.

Measuring bias is important as it helps identify flaws in quantitative forecasting methods or judgmental forecasts. It can, therefore, potentially help improve forecasts. Despite this, bias tends to be under-represented in the literature: many studies focus solely on measuring accuracy. Methods for assessing bias in single series are relatively well-known and well-researched, but for datasets containing thousands of observations for multiple series, the methodology for measuring and reporting bias is less obvious. We compare alternative approaches against a number of criteria when rolling-origin point forecasts are available for different forecasting methods and for multiple horizons over multiple series. We focus on relatively simple, yet interpretable and easy-to-implement metrics and visualization tools that are likely to be applicable in practice. To study the statistical properties of alternative measures we use theoretical concepts and simulation experiments based on artificial data with predetermined features. We describe the difference between mean and median bias, describe the connection between metrics for accuracy and bias, provide suitable bias measures depending on the loss function used to optimise forecasts, and suggest which measures for accuracy should be used to accompany bias indicators. We propose several new measures and provide our recommendations on how to evaluate forecast bias across multiple series.

A random variable X that is base b Benford will not in general be base c Benford when c is not equal to b. This paper builds on two of my earlier papers and is an attempt to cast some light on the issue of base dependence. Following some introductory material, the “Benford spectrum” of a positive random variable is introduced and known analytic results about Benford spectra are summarized. Some standard machinery for a “Benford analysis” is introduced and combined with my method of “seed functions” to yield tools to analyze the base c Benford properties of a base b Benford random variable. Examples are generated by applying these general methods to several families of Benford random variables. Berger and Hill's concept of “base-invariant significant digits” is discussed. Some potential extensions are sketched.

In this note, I analyze the data generated by M. Fodje's (2013) simulation programs "epr-simple" and "epr-clocked". They were written in Python and published on Github. Inspection of the program descriptions shows that they make use of the detection-loophole and the coincidence-loophole respectively. I evaluate them with appropriate modified Bell-CHSH type inequalities: the Larsson detection-loophole adjusted CHSH, and the Larsson-Gill coincidence-loophole adjusted CHSH (NB: its correctness is conjecture, we do not have proof). The experimental efficiencies turn out to be approximately eta = 81% (close to optimal) and gamma = 55% (far from optimal). The observed values of CHSH are, as they should be, within the appropriately adjusted bounds. Fodjes' detection-loophole model turns out to be very, very close to Pearle's famous 1970 model, so the efficiency is close to optimal. The model has the same defect as Pearle's: the joint detection rates exhibit signaling. Fodje's coincidence-loophole model is actually a clever modification of his detection-loophole model. Because of this, however, it cannot lead to optimal efficiency.

The paper is concerned with a class of stochastic differential equations in infinite dimensional Hilbert space with random coefficients driven by Teugel's martingales which are more general processes. and its optimal control problem. Here Teugels martingales are a family of pairwise strongly orthonormal martingales associated with L\'{e}vy processes (see Nualart and Schoutens). There are three major ingredients. The first is to prove the existence and uniqueness of the solutions by continuous dependence theorem of solutions combining with the parameter extension method. The second is to establish the stochastic maximum principle and verification theorem for our optimal control problem by the classicconvex variation method and dual technique.The third is to represent an example of a Cauchy problem for a controlled stochastic partial differential equation driven by Teugels martingales which our theoretical results can solve.

It is quite easy to stochastically distort an original count variable to obtain a new count variable with relatively more variability than in the original variable. Many popular overdispersion models (variance greater than mean) can indeed be obtained by mixtures, compounding or randomlystopped sums. There is no analogous stochastic mechanism for the construction of underdispersed count variables (variance less than mean), starting from an original count distribution of interest. This work proposes a generic method to stochastically distort an original count variable to obtain a new count variable with relatively less variability than in the original variable. The proposed mechanism, termed condensation, attracts probability masses from the quantiles in the tails of the original distribution and redirect them toward quantiles around the expected value. If the original distribution can be simulated, then the simulation of variates from a condensed distribution is straightforward. Moreover, condensed distributions have a simple mean-parametrization, a characteristic useful in a count regression context. An application to the negative binomial distribution resulted in a distribution allowing under, equi and overdispersion. In addition to graphical insights, fields of applications of special cases of condensed Poisson and condensed negative binomial distributions were pointed out as an indication of the potential of condensation for a flexible analysis of count data

In this paper, we focus on variational Bayesian learning deterministic optimization methods for inference in biparametric exponential models where the parameters follow semiparametric regression structures. This combination of data models and algorithms contributes to solving real-world problems and reduces the computation time. This allows both the rapid exploration of many data models and the accurate estimation of the mean and variance functions through the connection between generalized linear models and graph theory. A simulation study was carried out to assess the performance of the deterministic approximation. Finally, herein, we present an application using macroeconomic data to emphasize the benefits of the proposed approach.

This article examines the impact of some system parameters on an industrial system composed of two dissimilar parallel units with one repairman. The active unit may fail due to essential factors like aging or deteriorating, or exterior phenomena such as Poisson shocks that occur at various time periods. Whenever the value of a shock is larger than the specified threshold of the active unit, the active unit will fail. The article assumes that the repairman has the right to take any of two decisions at the beginning of the system operation: either a takes a vacation if the two units work in a normal way, or stay in the system to monitor the system until the first system failure. In case of having a failure in any of the two units during the absence of the repairman, the failing unit will have to wait until the repairman is called back to work. We suppose that the value of every shock is assumed to be i.i.d. with some known distribution. The length of the repairman’s vacation, repair time, and recall time are arbitrary distributions. Various reliability measures have been calculated by the supplementary variable technique and the Markov’s vector process theory. At last, numerical computation and graphical analysis have been given for a particular case to validate the derived indices.

Most existing flexible count regression models allow only approximate inference. This work proposes a new framework to provide an exact and flexible alternative for modeling and simulating count data with various types of dispersion (equi-, under- and overdispersion). The new method, referred as “balanced discretization”, consists in discretizing continuous probability distributions while preserving expectations. It is easy to generate pseudo random variates from the resulting balanced discrete distribution since it has a simple stochastic representation in terms of the continuous distribution. For illustrative purposes, we have developed the family of balanced discrete gamma distributions which can model equi-, under- and overdispersed count data. This family of count distributions is appropriate for building flexible count regressionmodels because the expectation of the distribution has a simple expression in terms of the parameters of the distribution. Using the Jensen–Shannon divergence measure, we have shown that under equidispersion restriction, the family of balanced discrete gamma distributions is similar to the Poisson distribution. Based on this, we conjecture that while covering all types of dispersion, a count regression model based on the balanced discrete gamma distribution will allow recovering a near Poisson distribution model fit when the data is Poisson distributed.

Study objective Since December 2019, the coronavirus disease (COVID-19) pandemic has caused over a million deaths and resulted in adverse socio-economic impacts worldwide. However, predictability and prognostication of clinical features vary among different populations. Methods We search PubMed, EMBASE, Cochrane Library, Google Scholar, and WHO Global Health Library from December 2019 to April 2020 for studies reporting the risk factors, clinical features, and outcomes. The random-effect models for transformed prevalence (single-arm) or bivariate random-effect models (sensitivity and specificity) for correlated performance indicators. Results Among the 189 included studies representing 53,659 patients, the most sensitive predictor for COVID-19 infection was fever in adults (83%, 95% confidence interval [CI]:73–90%), and the most specific predictor was fatigue (96%, 95% CI: 80–99%). Fever was the most sensitive symptom in predicting the severity (89%, 95% CI:83–92%), followed by cough (71%, 95% CI:63–78%). The most specific predictor of severe COVID-19 was a chronic obstructive pulmonary disease (99%, 95% CI:98–99%). The stage of the outbreak and age significantly affect the prevalence of fever, fatigue, cough, and dyspnea. Fever, cough, fatigue, hypertension, and diabetes mellitus combined have a 3.06 positive likelihood ratio (PLR) and a 0.59 negative likelihood ratio (NLR) in the diagnosis. Additionally, fever, cough, sputum production, myalgia, fatigue, and dyspnea combined have a 10.44 PLR and a 0.16 NLR in predicting severe COVID-19. Conclusions Understanding the different distribution of predictors essential for screening potential COVID-19 infection and severe outcomes and the combination of symptoms could improve the pre-test probability.

A new discrete distribution for count data called extended biparametric Waring (EBW) distribution is developed. Its name is related to the fact that, in a specific configuration of its parameters, it can be seen as a biparametric version of the univariate generalized Waring (UGW) distribution, a well-known model for the variance decomposition into three components: randomness, liability and proneness. Unlike the UGW distribution, the EBW can model both overdispersed and underdispersed data sets. In fact, the EBW distribution is a particular case of a UWG distribution when its first parameter is positive; otherwise, it is a particular case of a Complex Triparametric Pearson (CTP) distribution. Hence, this new model inherits most of their properties and, moreover, it helps to solve the identification problem in the variance components of the UGW model. We compare the EBW with the UGW by a simulation study, but also with other over and underdispersed distributions through the Kullback-Leibler divergence. Additionally, we have carried out a simulation study in order to analyse the properties of the maximum likelihood parameter estimates. Finally, some application examples are included which show that the proposed model provides similar or even better results than other models, but with fewer parameters.

The road traffic injuries risk factors such as driving offenses and average speed are concerns for health organizations to reduce the number of injuries. Without any comprehensive view of each road, one cannot decide about the effective policy. In this manner, the data-driven policy will help to improve and assess the decisions. The count data near the road of two airports is surveyed for investigating the time-varying speed zones. The descriptive statistics, ANOVA, and functional data analysis were used. The hourly data of traffic counts for four different locations at the entrance of the two airports, international and domestics, were collected for one the year 2018 to 2019.The hourly pattern of driving offenses for each road was assessed and the to and from airport roads had different peaks (<0.05). The hour, weekdays, type of airport, direction and their interactions were statistically significant (<0.05) for the chance of driving offenses. The speed average during the day was statistically different (<0.5) by the number of different types of vehicles. The traffic count data is a great resource for decision making in safe driving subjects such as driving offenses. With functional data analysis, we can analyze them to get the most of the characteristics of this data. The airports are public places with high traffic demand in all countries that yields the different pattern of traffic transportation, therefore we extract the factors that affect the driving offenses. Finally, we conclude that conducting a time-varying speed zone near the airports seems vital.

Maximum likelihood estimation (MLE) is a popular method for parameter estimation in both applied probability and statistics but MLE cannot solve the problem of incomplete data or hidden data because it is impossible to maximize likelihood function from hidden data. Expectation maximum (EM) algorithm is a powerful mathematical tool for solving this problem if there is a relationship between hidden data and observed data. Such hinting relationship is specified by a mapping from hidden data to observed data or by a joint probability between hidden data and observed data. In other words, the relationship helps us know hidden data by surveying observed data. The essential ideology of EM is to maximize the expectation of likelihood function over observed data based on the hinting relationship instead of maximizing directly the likelihood function of hidden data. Pioneers in EM algorithm proved its convergence. As a result, EM algorithm produces parameter estimators as well as MLE does. This tutorial aims to provide explanations of EM algorithm in order to help researchers comprehend it. Moreover some improvements of EM algorithm are also proposed in the tutorial such as combination of EM and third-order convergence Newton-Raphson process, combination of EM and gradient descent method, and combination of EM and particle swarm optimization (PSO) algorithm.

Expectation maximization (EM) algorithm is a powerful mathematical tool for estimating statistical parameter when data sample contains hidden part and observed part. EM is applied to learn finite mixture model in which the whole distribution of observed variable is average sum of partial distributions. Coverage ratio of every partial distribution is specified by the probability of hidden variable. An application of mixture model is soft clustering in which cluster is modeled by hidden variable whereas each data point can be assigned to more than one cluster and degree of such assignment is represented by the probability of hidden variable. However, such probability in traditional mixture model is simplified as a parameter, which can cause loss of valuable information. Therefore, in this research I propose a so-called conditional mixture model (CMM) in which the probability of hidden variable is modeled as a full probabilistic density function (PDF) that owns individual parameter. CMM aims to extend mixture model. I also propose an application of CMM which is called adaptive regressive model (ARM). Traditional regression model is effective when data sample is scattered equally. If data points are grouped into clusters, regression model tries to learn a unified regression function which goes through all data points. Obviously, such unified function is not effective to evaluate response variable based on grouped data points. The concept “adaptive” of ARM means that ARM solves the ineffectiveness problem by selecting the best cluster of data points firstly and then evaluating response variable within such best cluster. In order words, ARM reduces estimation space of regression model so as to gain high accuracy in calculation.

We construct admissible circulant Laplacian matrix functions as generators for strictly increasing random walks on the integer line. These Laplacian matrix functions refer to a certain class of Bernstein functions. The approach has connections with biased walks on digraphs. Within this framework, we introduce a space-time generalization of the Poisson process as a strictly increasing walk with discrete Mittag-Leffler jumps subordinated to a (continuous-time) fractional Poisson process. We call this process ‘space-time Mittag-Leffler process’. We derive explicit formulae for the state probabilities which solve a Cauchy problem with a Kolmogorov-Feller (forward) difference-differential equation of general fractional type. We analyze a “well-scaled” diffusion limit and obtain a Cauchy problem with a space-time convolution equation involving Mittag-Leffler densities. We deduce in this limit the ‘state density kernel’ solving this Cauchy problem. It turns out that the diffusion limit exhibits connections to Prabhakar general fractional calculus. We also analyze in this way a generalization of the space-time fractional Mittag-Leffler process. The approach of construction of good Laplacian generator functions has a large potential in applications of space-time generalizations of the Poisson process and in the field of continuous-time random walks on digraphs.

For bounded unit interval, we propose a new Kumaraswamy generalized (G) family of distributions from a new generator which could be an alternate to the Kumaraswamy-G family proposed earlier by Cordeiro and de-Castro in 2011. This new generator can also be used to develop alternate G-classes such as beta-G, McDonald-G, Topp-Leone-G, Marshall-Olkin-G and Transmuted-G for bounded unit interval. Some mathematical properties of this new family are obtained and maximum likelihood method is used for estimating the family parameters. We investigate the properties of one special model called a new Kumaraswamy-Weibull (NKwW) distribution. Parameter estimation is dealt and maximum likelihood estimators are assessed through simulation study. Two real life data sets are analyzed to illustrate the importance and flexibility of this distribution. In fact, this model outperforms some generalized Weibull models such as the Kumaraswamy-Weibull, McDonald-Weibull, beta-Weibull, exponentiated-generalized Weibull, gamma-Weibull, odd log-logistic-Weibull, Marshall-Olkin-Weibull, transmuted-Weibull, exponentiated-Weibull and Weibull distributions when applied to these data sets. The bivariate extension of the family is proposed and the estimation of parameters is given. The usefulness of the bivariate NKwW model is illustrated empirically by means of a real-life data set.

Conditional heteroskedastic financial time series are commonly modelled by (G)ARCH processes. ARCH(1) and GARCH were recently established in C[0,1] and L^2[0,1]. This article provides sufficient conditions for the existence of strictly stationary solutions, weak dependence and finite moments of (G)ARCH processes for any order in C[0,1] and L^p[0,1]. It deduces explicit asymptotic upper bounds of estimation errors for the shift term, the complete (G)ARCH operators and the projections of ARCH operators on finite-dimensional subspaces. The operator estimaton is based on Yule-Walker equations, and estimating the GARCH operators also involves a result estimating operators in invertible linear processes being valid beyond the scope of (G)ARCH. Moreover, our results regarding (G)ARCH can be transferred to functional AR(MA).