Four-dimensional variational data assimilation for Doppler radar wind data

All forecast models, whether they represent the state of the weather, the spread of a disease, or levels of economic activity, contain unknown parameters. These parameters may be the model's initial conditions, its boundary conditions, or other tunable parameters which have to be determined. Four dimensional variational data assimilation (4D-Var) is a method of estimating this set of parameters by optimizing the fit between the solution of the model and a set of observations which the model is meant to predict.Although the method of 4D-Var described in this paper is not restricted to any particular system, the application described here has a numerical weather prediction (NWP) model at its core, and the parameters to be determined are the initial conditions of the model.The purpose of this paper is to give a review covering assimilation of Doppler radar wind data into a NWP model. Some associated problems, such as sensitivity to small variations in the initial conditions or due to small changes in the background variables, and biases due to nonlinearity are also studied.     

Stochastic models for first-order kinetics of biochemical oxygen demand with random initial conditions, inputs, and coefficients

Biochemical oxygen demand (BOD) is a parameter of prime importance in surface water pollution studies and in the design and operation of waste-water treatment plants. A general, stochastic analytical model (denoted S1) is developed for the temporal expectation and (heteroscedastic) variance of first-order BOD kinetics. The model is obtained by integrating the moment equation, which is derived from the mathematical theory of stochastic differential equations. This model takes into account random initial conditions, random inputs, and random coefficients, which appear in the model formulation as initial condition (σ_O²), input (σ_l²), and coefficient (σ_c²) variance parameters, respectively. By constraining these three variance parameters to either vanish or to be nonnegative, model S1 is allowed (under appropriate combinations of the constraints) to split into six stochastic “submodels” (denoted S2 to S7), with each of these submodels being a particular case of the general model. Model S1 also degenerates to the deterministic model (denoted D) when each of the variance parameters vanish. The deterministic parameters (i.e., the rate coefficient and the ultimate BOD) and the stochastic variance parameters of the seven models are estimated on sets of replicated BOD data using the maximum likelihood principle. In this study, two (S5 and S7) of these seven stochastic models are found to be appropriate for BOD. The stochastic input (S5) model (i.e., null initial condition and coefficient variance parameters) shows the best prediction capabilities, while the next best is the stochastic initial condition (S7) model (i.e., null input and coefficient variance parameters).     

Bayesian Inference for the RC(m) Association Model

Describing the structure in a two-way contingency table in terms of an RC(m) association model, we are concerned with the computation of posterior distributions of the model parameters using prior distributions which take into account the nonlinear restrictions of the model. We are further involved with the determination of the order of association m, based on Bayesian arguments. Using projection methods, a prior distribution over the parameters of the simpler RC(m) model is induced from a prior of the parameters of the saturated model. The fit of the assumed RC(m) model is evaluated using the posterior distribution of its distance from the full model. Our methods are illustrated with a popular dataset.     

Generalised linear model trees with global additive effects

Model-based trees are used to find subgroups in data which differ with respect to model parameters. In some applications it is natural to keep some parameters fixed globally for all observations while asking if and how other parameters vary across subgroups. Existing implementations of model-based trees can only deal with the scenario where all parameters depend on the subgroups. We propose partially additive linear model trees (PALM trees) as an extension of (generalised) linear model trees (LM and GLM trees, respectively), in which the model parameters are specified a priori to be estimated either globally from all observations or locally from the observations within the subgroups determined by the tree. Simulations show that the method has high power for detecting subgroups in the presence of global effects and reliably recovers the true parameters. Furthermore, treatment–subgroup differences are detected in an empirical application of the method to data from a mathematics exam: the PALM tree is able to detect a small subgroup of students that had a disadvantage in an exam with two versions while adjusting for overall ability effects.

     

Estimating Functions for Nonlinear Time Series Models

This paper discusses the problem of estimation for two classes of nonlinear models, namely random coefficient autoregressive (RCA) and autoregressive conditional heteroskedasticity (ARCH) models. For the RCA model, first assuming that the nuisance parameters are known we construct an estimator for parameters of interest based on Godambe's asymptotically optimal estimating function. Then, using the conditional least squares (CLS) estimator given by Tjøstheim (1986, Stochastic Process. Appl., 21, 251–273) and classical moment estimators for the nuisance parameters, we propose an estimated version of this estimator. These results are extended to the case of vector parameter. Next, we turn to discuss the problem of estimating the ARCH model with unknown parameter vector. We construct an estimator for parameters of interest based on Godambe's optimal estimator allowing that a part of the estimator depends on unknown parameters. Then, substituting the CLS estimators for the unknown parameters, the estimated version is proposed. Comparisons between the CLS and estimated optimal estimator of the RCA model and between the CLS and estimated version of the ARCH model are given via simulation studies.     

Generalizing the Pareto to the log-Pareto model and statistical inference

In this article we introduce a full-fledged statistical model of log-Pareto distribution functions (dfs) parametrized by two shape parameters and a scale parameter. Pareto dfs can be regained in the limit by varying parameters of log-Pareto dfs, whence the log-Pareto model can be regarded as an extension of the Pareto model. Log-Pareto dfs are first of all obtained by means of exponential transformations of Pareto dfs. We also indicate an iterated application of such a procedure. A class of generalized log-Pareto dfs is considered as well. In addition, power-pot (p-pot) stable dfs – related to p-max stable dfs – are introduced and log-Pareto dfs are identified as special cases. A modification of a quick (systematic) estimator is proposed as an initial estimator for the numerical computation of the maximum likelihood estimator (MLE) in the 3-parameter model.      

Fast derivatives of likelihood functionals for ODE based models using adjoint-state method

We consider time series data modeled by ordinary differential equations (ODEs), widespread models in physics, chemistry, biology and science in general. The sensitivity analysis of such dynamical systems usually requires calculation of various derivatives with respect to the model parameters. We employ the adjoint state method (ASM) for efficient computation of the first and the second derivatives of likelihood functionals constrained by ODEs with respect to the parameters of the underlying ODE model. Essentially, the gradient can be computed with a cost (measured by model evaluations) that is independent of the number of the ODE model parameters and the Hessian with a linear cost in the number of the parameters instead of the quadratic one. The sensitivity analysis becomes feasible even if the parametric space is high-dimensional. The main contributions are derivation and rigorous analysis of the ASM in the statistical context, when the discrete data are coupled with the continuous ODE model. Further, we present a highly optimized implementation of the results and its benchmarks on a number of problems. The results are directly applicable in (e.g.) maximum-likelihood estimation or Bayesian sampling of ODE based statistical models, allowing for faster, more stable estimation of parameters of the underlying ODE model.     

A parameterized approach to modeling and forecasting mortality

A new method is proposed of constructing mortality forecasts. This parameterized approach utilizes Generalized Linear Models (GLMs), based on heteroscedastic Poisson (non-additive) error structures, and using an orthonormal polynomial design matrix. Principal Component (PC) analysis is then applied to the cross-sectional fitted parameters. The produced model can be viewed either as a one-factor parameterized model where the time series are the fitted parameters, or as a principal component model, namely a log-bilinear hierarchical statistical association model of Goodman [Goodman, L.A., 1991. Measures, models, and graphical displays in the analysis of cross-classified data. J. Amer. Statist. Assoc. 86(416), 1085-1111] or equivalently as a generalized Lee-Carter model with p interaction terms. Mortality forecasts are obtained by applying dynamic linear regression models to the PCs. Two applications are presented: Sweden (1751-2006) and Greece (1957-2006).     

Bayesian Filtering for Jump-Diffusions With Application to Stochastic Volatility

In this article, the problem of sequentially learning parameters governing discretely observed jump-diffusions is explored. The estimation framework involves the introduction of latent points between every pair of observations to allow a sufficiently accurate Euler–Maruyama approximation of the underlying (but unavailable) transition densities. Particle filtering algorithms are then implemented to sample the posterior distribution of the latent data and the model parameters online. The methodology is applied to the estimation of parameters governing a stochastic volatility (SV) model with jumps. As well as using S&P 500 Index data, a simulation study is provided. Supplemental materials for this article are available online.     

A Recursive Algorithm to Restore Images Based on Robust Estimation of NSHP Autoregressive Models

The objective of this article is to present a new image restoration algorithm. First, each pixel in the image is classified into k categories. Then we assume that the gray levels in each category follow a nonsymmetric half-plane (NSHP) autoregressive model. Robust estimation of the parameters of the model is considered to attenuate the effect of the image contamination on the parameters. In each iteration we will construct a new image using a robustified version of the residuals. The introduction of the classification techniques as a first step of the algorithm reduces considerably the number of parameters to estimate. Hence, the computational time is also reduced because the robust estimations of the parameters are solutions of nonlinear system of equations. Some applications are presented to real synthetic aperture radar (SAR) images to illustrate how our algorithm restores an image in practice.     

Estimating the parameters of Weibull distribution and the acceleration factor from hybrid partially accelerated life test

This article considers the estimation of parameters of Weibull distribution based on hybrid censored data. The parameters are estimated by the maximum likelihood method under step-stress partially accelerated test model. The maximum likelihood estimates (MLEs) of the unknown parameters are obtained by Newton–Raphson algorithm. Also, the approximate Fisher information matrix is obtained for constructing asymptotic confidence bounds for the model parameters. The biases and mean square errors of the maximum likelihood estimators are computed to assess their performances through a Monte Carlo simulation study.     

Probabilistic nonlinear analysis of CFR dams by MCS using Response Surface Method

This paper presents the probabilistic analysis of concrete-faced rockfill (CFR) dams according to the Monte Carlo Simulation (MCS) results which are obtained through the Response Surface Method (RSM). ANSYS finite element program is used to get displacement and principal stress components. First of all, some parametric studies are performed according to the simple and representative finite element model of dam body to obtain the optimum approximate model. Secondly, a sensitivity analysis is performed to get the most effective parameters on dam response. Then, RSM is used to obtain the approximate function through the selected parameters. After the performed analyses, star experimental design with quadratic function without mixed terms according to the k = 1 is determined as the most appropriate model. Finally, dam-foundation-reservoir interaction finite element model is constituted and probabilistic analyses are performed with MCS using the selected parameters, sampling method, function and arbitrary factor under gravity load for empty and full reservoir conditions. Geometrically and materially nonlinearity are considered in the analysis of dam-foundation-reservoir interaction system. Reservoir water is modeled by fluid finite elements based on the Lagrangian approach. Structural connections are modeled as welded contact and friction contact based on Coulomb’s friction law. Probabilistic displacements and stresses are presented and compared with deterministic results.     

Extracting compact fuzzy rules for nonlinear system modeling using subtractive clustering,GA and unscented filter

This paper presents a two stage procedure for building optimal fuzzy model from data for nonlinear dynamical systems. Both stages are embedded into Genetic Algorithm (GA) and in the first stage emphasis is placed on structural optimization by assigning a suitable fitness to each individual member of population in a canonical GA. These individuals represent coded information about the structure of the model (number of antecedents and rules). This information is consequently utilized by subtractive clustering to partition the input space and construct a compact fuzzy rule base. In the second stage, Unscented Filter (UF) is employed for optimization of model parameters, that is, parameters of the input–output Membership Functions (MFs).     

A note on reduction of the number of parameters in linear statistical models

In certain settings the mean response is modeled by a linear model using a large number of parameters. Sometimes it is desirable to reduce the number of parameters prior to conducting the experiment and prior to the actual statistical analysis. Essentially, it means to formulate a simpler approximate model to the original “ideal” one. The goal is to find conditions (on the model matrix and covariance matrix) under which the reduction does not influence essentially the data fit. Here we try to develop such conditions in regular linear model without and with linear restraints. We emphasize that these conditions are independent of observed data.     

The Shadow Prior

In this article we consider posterior simulation in models with constrained parameter or sampling spaces. Constraints on the support of sampling and prior distributions give rise to a normalization constant in the complete conditional posterior distribution for the (hyper-) parameters of the respective distribution, complicating posterior simulation.

To mitigate the problem of evaluating normalization constants, we propose a computational approach based on model augmentation. We include an additional level in the probability model to separate the (hyper-) parameter from the constrained probability model, and we refer to this additional level in the probability model as a shadow prior. This approach can significantly reduce the overall computational burden if the original (hyper-) prior includes a complicated structure, but a simple form is chosen for the shadow prior, for example, if the original prior includes a mixture model or multivariate distribution, and the shadow prior defines a set of shadow parameters that are iid given the (hyper-) parameters. Although introducing the shadow prior changes the posterior inference on the original parameters, we argue that by appropriate choices of the shadow prior, the change is minimal and posterior simulation in the augmented probability model provides a meaningful approximation to the desired inference. Data used in this article are available online.     

Parameter Estimation of the Disk in a Floating Caliper Disk Brake Model with respect to Squeal

Noise from a vehicle is always a concern for any automotive industry looking for passenger comfort. This also holds for the different types of brake noise, in particular for squeal, which is a source of discomfort both to passengers and passers‐by. Intensive research on low frequency squeal (noise between 1‐5 kHz) has been carried out. A model of the floating caliper disk brake has been recently proposed by the authors with the aim to predict the onset of squeal. A flutter type instability resulting from nonconservative restoring forces is assumed to be the reason behind squeal in this model. The problem of validating the model against experimental observation lies in the fact that all the system parameters included in the model, especially the disk, are to be carefully chosen. In this work some parameters of the disk are estimated and a method is suggested to estimate these parameters in a systematic way. For that purpose, the values are to be accurately selected by experiments and modal updating technique. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Using ARMA models to forecast workpiece roundness error in a turning operation

The paper describes the methodology for developing autoregressive moving average (ARMA) models to represent the workpiece roundness error in the machine taper turning process. The method employs a two stage approach in the determination of the AR and MA parameters of the ARMA model. It first calculates the parameters of the equivalent autoregressive model of the process, and then derives the AR and MA parameters of the ARMA model. Akaike's Information Criterion (AIC) is used to find the appropriate orders m and n of the AR and MA polynomials respectively. Recursive algorithms are developed for the on-line implementation on a laboratory turning machine. Evaluation of the effectiveness of using ARMA models in error forecasting is made using three time series obtained from the experimental machine. Analysis shows that ARMA(3,2) with forgetting factor of 0.95 gives acceptable results for this lathe turning machine.     

Multicriteria sorting using a valued indifference relation under a preference disaggregation paradigm

Most multicriteria decision methods need the definition of a significant amount of preferential information from a decision agent. The preference disaggregation analysis paradigm infers the model’s parameter values from holistic judgments provided by a decision agent. Here, a new method for inferring the parameters of a fuzzy outranking model for multicriteria sorting is proposed. This approach allows us to use most of the preferential information contained in a reference set. The central idea is to characterize the quality of the model by measuring discrepancies and concordances amongst (i) the preference relations derived from the outranking model, and (ii) the preferential information contained in the reference set. The model’s parameters are inferred from a multiobjective optimization problem, according to some additional preferential information from a decision agent. Once the model has been fitted, sorting decisions about new objects are performed by using a fuzzy indifference relation. This proposal performs very well in some examples.     

A genetic algorithm for optimal operating parameters of VMI system in a two-echelon supply chain

This paper deals with the operational issues of a two-echelon single vendor–multiple buyers supply chain (TSVMBSC) model under vendor managed inventory (VMI) mode of operation. The operational parameters to the above model are: sales quantity and sales price that determine the channel profit of the supply chain, and contract price between the vendor and the buyer, which depends upon the understanding between the partners on their revenue sharing. In order to find out the optimal sales quantity for each buyer in TSVMBSC problem, a mathematical model is formulated. Optimal sales price and acceptable contract price at different revenue share are subsequently derived with the optimal sales quantity. A genetic algorithm (GA) based heuristic is proposed to solve this TSVMBSC problem, which belongs to nonlinear integer programming problem (NIP). The proposed methodology is evaluated for its solution quality. Furthermore, the robustness of the model with its parameters, which fluctuate frequently and are sensitive to operational features, is analysed.     

Statistical inference on parametric part for partially linear single-index model

Statistical inference on parametric part for the partially linear single-index model (PLSIM) is considered in this paper. A profile least-squares technique for estimating the parametric part is proposed and the asymptotic normality of the profile least-squares estimator is given. Based on the estimator, a generalized likelihood ratio (GLR) test is proposed to test whether parameters on linear part for the model is under a contain linear restricted condition. Under the null model, the proposed GLR statistic follows asymptotically the χ²-distribution with the scale constant and degree of freedom independent of the nuisance parameters, known as Wilks phenomenon. Both simulated and real data examples are used to illustrate our proposed methods.