Recursive parameter estimation: convergence

We consider estimation procedures which are recursive in the sense that each successive estimator is obtained from the previous one by a simple adjustment. We propose a wide class of recursive estimation procedures for the general statistical model and study convergence.      

Error structures and parameter estimation

The error calculus based on the theory of Dirichlet forms is an extension of Gauss' approach to error propagation. The aim of this paper is to derive error structures from measurements. The links with Fisher's information lay the foundations of a strong connection with experiment. Here we show that this connection behaves well towards changes of variables and is related to the theory of asymptotic statistics. Finally the study of products permits one to lay the foundation of an infinite dimensional empirical error calculus. To cite this article: N. Bouleau, C. Chorro, C. R. Acad. Sci. Paris, Ser. I 338 (2004).     

On optimal parameter estimation in credibility

The present paper is concerned with optimal estimation of the 1st and 2nd order structural moments appearing in credibility formulas. In a recent paper De Vylder has treated the problem in the case of multinormal conditional distributions under quite restrictive assumptions. He minimizes, within a certain restricted class of unbiased estimators, the variance (or the sum of variances if the estimand is a matrix) and next replaces all structural moments (up to fourth order) in the solution by estimates based on the data. This paper is an attempt to simplify the method and extend it so as to make it applicable in more general situations. By suitable choice of a (sufficient) set of statistics and a suitable parametrization, the powerful theory of estimation in linear models can be employed, which makes cumbersome minimization procedures superfluous. The theory is applied to the cases with binomial. Poisson, compound Poisson, and multinormal conditional distributions. Some simulation studies have been performed to assess the performance of the estimators.     

Recursive parameter estimation: asymptotic expansion

This paper is concerned with the asymptotic behaviour of estimation procedures which are recursive in the sense that each successive estimator is obtained from the previous one by a simple adjustment. The results of the paper can be used to determine the form of the recursive procedure which is expected to have the same asymptotic properties as the corresponding non-recursive one defined as a solution of the corresponding estimating equation. Several examples are given to illustrate the theory, including an application to estimation of parameters in exponential families of Markov processes.     

On parameter estimation for cusp-type signals

We consider the problem of parameter estimation by continuous time observations of a deterministic signal in white Gaussian noise. It is supposed that the signal has a cusp-type singularity. The properties of the maximum-likelihood and Bayesian estimators are described in the asymptotics of small noise. Special attention is paid to the problem of parameter estimation in the situation of misspecification in regularity, i.e., when the statistician supposes that the observed signal has this singularity, but the real signal is smooth. The rate and the asymptotic distribution of the maximum-likelihood estimator in this situation are described.     

Semiparametric estimation of the long-range parameter

We study two estimators of the long-range parameter of a covariance stationary linear process. We show that one of the estimators achieve the optimal semiparametric rate of convergence, whereas the other has a rate of convergence as close as desired to the optimal rate. Moreover, we show that the estimators are asymptotically normal with a variance, which does not depend on any unknown parameter, smaller than others suggested in the literature. Finally, a small Monte Carlo study is included to illustrate the finite sample relative performance of our estimators compared to other suggested semiparametric estimators. More specifically, the Monte-Carlo experiment shows the superiority of the proposed estimators in terms of the Mean Squared Error. The first author research was funded by the Economic and Social Research Council (ESRC) reference number: R000238212. The second author research was funded by the Ministry of Education, Culture, Sports and Technology of Japan, reference number: 09CE2002 and B(2)10202202.     

A nonparametric estimation of the infection curve

Predicting the future course of an epidemic depends on being able to estimate the current numbers of infected individuals.However,while back-projection techniques allow reliable estimation of the numbers of infected individuals in the more distant past,they are less reliable in the recent past.We propose two new nonparametric methods to estimate the unobserved numbers of infected individuals in the recent past in an epidemic.The proposed methods are noniterative,easily computed and asymptotically normal wit...     

A parameter estimation method for model analysis

To solve a class of nonlinear parameter estimation problems, a method combining the regularized structured nonlinear total least norm (RSNTLN) method and parameter separation scheme is suggested. The method guarantees the convergence of parameters and has an advantages in reducing the residual norm over the use of RSNTLN only. Numerical experiments for two models appeared in signal processing show that the suggested method is more effective in obtaining solution and parameter with minimum residual norm.     

On sequential estimation of the shift parameter

The problem is considered of estimating the minimal sample size that guarantees the required accuracy, with confidence level fixed, of the estimate of the expectation. Translated fromStatisticheskie Metody Otsenivaniya i Proverki Gipotez, pp. 32–36, Perm, 1991.     

Optimal inputs and sensitivities for parameter estimation

Observed data for parameter estimation is often both difficult and expensive to obtain. Thus, when an experiment is conducted, the input to the system should be such that the sensitivity to the parameter being estimated is maximized. Evaluation of the optimal inputs and sensitivities requires the solution to a two-point boundary-value problem. The method of quasilinearization is then used for parameter estimation. Numerical results are given for a simple example utilizing both optimal and nonoptimal inputs. The results clearly show the advantages of utilizing an optimal input.     

Guaranteed nonlinear parameter estimation in knowledge-based models

Knowledge-based models are ubiquitous in pure and applied sciences. They often involve unknown parameters to be estimated from experimental data. This is usually much more difficult than for black-box models, only intended to mimic a given input–output behavior. The output of knowledge-based models is almost always nonlinear in their parameters, so that linear least squares cannot be used, and analytical solutions for the model equations are seldom available. Moreover, since the parameters have some physical meaning, it is not enough to find some numerical values of these quantities that are such that the model fits the data reasonably well. One would like, for instance, to make sure that the parameters to be estimated are identifiable. If this is not the case, all equivalent solutions should be provided. The uncertainty in the parameters resulting from the measurement noise and approximate nature of the model should also be characterized. This paper describes how guaranteed methods based on interval analysis may contribute to these tasks. Examples in linear and nonlinear compartmental modeling, widely used in biology, are provided.     

State and parameter estimation for nonlinear systems

State and parameter estimators are obtained for systems described by nonlinear evolution equations. Linear infinite dimensional observability theory together with a variety of fixed point theorems can be employed to obtain a finite time observer. Moreover, a nonlinear asymptotic observer is produced using stability results. The problem of joint state and parameter estimation is converted to the state estimation case, via an augmented state, so that these observer results can be utilised. Examples and remarks on the generality of the results are given.     

Tumor location and parameter estimation by thermography

In non-invasive thermal diagnostics, accurate correlations between the thermal image on skin surface and interior human physiology are often desired, which require general solutions for the bioheat equation. In this study an estimation methodology is presented to determine unknown thermophysical or geometrical parameters of a tumor region using the temperature profile on the skin surface that may be obtained by infrared thermography. To solve these inverse problems a second order finite difference scheme was implemented to solve the bioheat Pennes equation with mixed boundary conditions in two and three dimensions. Then, the Pattern Search algorithm was used to estimate the different parameters by minimizing a fitness function involving the temperature profiles obtained from simulated or clinical data to those obtained by the finite different scheme.     

On parameter estimation of threshold autoregressive models

This paper studies the threshold estimation of a TAR model when the underlying threshold parameter is a random variable. It is shown that the Bayesian estimator is consistent and its limit distribution is expressed in terms of a limit likelihood ratio. Furthermore, convergence of moments of the estimators is also established. The limit distribution can be computed via explicit simulations from which testing and inference for the threshold parameter can be conducted. The obtained results are illustrated with numerical simulations.     

Procedure for asymptotic state and parameter estimation of nonlinear distributed parameter bioreactors

Due to measurement problem in biological variables, such as the biomass concentration and its specific growth rate, the on-line measurement of these variables is not ensured and this presents a great difficulty, especially when we consider the control problem. This paper presents and analyses an approach for estimating these biological variables through an example of a nonlinear distributed parameter bioreactor. An orthogonal collocation method is applied in the distributed system to obtain a system of ordinary differential equations. A simulation study shows the feasibility and the robustness of the estimator used for this nonlinear process.     

A bayesian approach to binary response curve estimation

Summary The purpose of the present paper is to propose a practical procedure for the estimation of the binary response curve. The procedure is based on a model which approximates the response curve by a finely segmented piecewise constant function. To obtain a stable estimate we assume a prior distribution of the parameters of the model. The prior distribution has several parameters (hyper-parameters) which are chosen to minimize an information criterion ABIC. The procedure is applicable to data consisting of observations of a binary response variable and a single explanatory variable. The practical utility of the procedure is demonstrated by examples of applications to the dose response curve estimation, to the intensity function estimation of a point process and to the analysis of social survey data. The application of the procedure to the discriminant analysis is also briefly discussed. The Institute of statistical mathematics     

Learning networks in rainfall estimation

This paper utilizes Artificial Neural Networks (ANNs), standard Support Vector Regression (SVR), Least-Squares Support Vector Regression (LS-SVR), linear regression (LR) and a rain rate (RR) formula that meteorologists use, to estimate rainfall. A unique source of ground truth rainfall data is the Oklahoma Mesonet. With the advent of the WSR-88D network of radars data mining is feasible for this study. The reflectivity measurements from the radar are used as inputs for the techniques tested. LS-SVR generalizes better than ANNs, linear regression and a rain rate formula in rainfall estimation and for rainfall detection, SVR has a better performance than the other techniques.Mathematics Subject Classification: 30C40     

The best parameter subset using the Chebychev curve fitting criterion

The Chebychev (also Minimax andL _∞ Norm) criterion has been widely studied as a method for curve fitting. Published computer codes are available to obtain the optimal parameter estimates to fit a linear function to a set of given points under the Chebychev criterion. The purpose of this paper is to study procedures for obtaining the best subset ofq parameters from a given set ofm parameters whereq is less-than-or-equal-tom.