Statistical inferences for varying-coefficint models based on locally weighted regression technique

1. IntroductionIn recent y6ars, some progress has been made towards increaJsing the flexibility of linearregression models. One of the obvious extensions in this direction is the sthcalled varyingcoefficient regression models in which the regressiOn funtions are llnear in the regressors,but their coefficients are allowed to change with the value of another factor. Specificallysuppose that we have a response variable Y and regressors X1 t X2,' t Xv as well as anothervariable V. A varyingco…     

Commodity and resource ETF trading patterns during the financial crisis

This study estimates the parameters of a power law fit of the distribution of log returns of exchange traded funds (ETFs) before, during, and after the recent financial crisis. It is found, that there is considerable variation both between ETFs and between calm and turbulent phases. Exponents of the daily log return distribution are estimated to lie mostly between 3.0 and 5.0 depending on the ETF. In minute‐by‐minute, trading data much lower power law exponents have been found concentrating between 3.0 and 4.0 and sometimes dropping to values close to or below 3.0. Further, there is evidence for changes in the distribution during times of turbulence (value of the exponent, improvement in the goodness of fit measures of the distribution). It can be hypothesized that effects such as, infinite variance (for α < 3) or changes in the form of the distribution can occur, in turn affecting the predictability of the system which has implications for the possibility to control or regulate financial markets under such conditions. © 2014 Wiley Periodicals, Inc. Complexity 21: 73–83, 2016     

On Markov chain Monte Carlo Algorithms for Computing Conditional Expectations Based on Sufficient Statistics

Much work has focused on developing exact tests for the analysis of discrete data using log linear or logistic regression models. A parametric model is tested for a dataset by conditioning on the value of a sufficient statistic and determining the probability of obtaining another dataset as extreme or more extreme relative to the general model, where extremeness is determined by the value of a test statistic such as the chi-square or the log-likelihood ratio. Exact determination of these probabilities can be infeasible for high dimensional problems, and asymptotic approximations to them are often inaccurate when there are small data entries and/or there are many nuisance parameters. In these cases Monte Carlo methods can be used to estimate exact probabilities by randomly generating datasets (tables) that match the sufficient statistic of the original table. However, naive Monte Carlo methods produce tables that are usually far from matching the sufficient statistic. The Markov chain Monte Carlo method used in this work (the regression/attraction approach) uses attraction to concentrate the distribution around the set of tables that match the sufficient statistic, and uses regression to take advantage of information in tables that “almost” match. It is also more general than others in that it does not require the sufficient statistic to be linear, and it can be adapted to problems involving continuous variables. The method is applied to several high dimensional settings including four-way tables with a model of no four-way interaction, and a table of continuous data based on beta distributions. It is powerful enough to deal with the difficult problem of four-way tables and flexible enough to handle continuous data with a nonlinear sufficient statistic.     

Random rates of growth and return: introducing the expo‐normal distribution

Expected values and standard deviations of the geometric means of independent positive random variables are useful indicators of the long‐term profitability of an investment, or survival of a biological population. Often these quantities cannot be evaluated in a closed form, or even if they can, there may be a choice between several probability models for the ‘annual’ growth factors. This paper formulates approximations for geometric means and standard deviations. It evaluates their performance and compares the best of them with the exact values for selected probability models of the annual factors. Among these is a new model for annual log‐returns, called the expo‐normal law. This is the law of log X, where X has a normal law conditioned on X>0. Its properties are developed in some detail. It is found that for the ranges of annual means and standard deviations typically encountered in financial applications, the longer horizon values depend little on the choice of probability model, and that, where possible, exact evaluation is computationally simpler than using approximations. Copyright © 2002 John Wiley & Sons, Ltd.     

Power Divergence Family of Tests for Categorical Time Series Models

A fundamental issue that arises after fitting a regression model is that of testing the goodness of the fit. Our work brings together the power divergence family of goodness of fit tests and regression models for categorical time series. We show that under some reasonable assumptions, the asymptotic distribution of the power divergence family of goodness of fit tests converges to a normal random variable. This fact introduces a novel method for carrying out goodness of fit tests about a regression model for categorical time series. We couple the theory with some empirical results.     

Estimating functions for conditional inference: Many nuisance parameter case

When there is a complete sufficient statistic for the nuisance parameter which depends on the parameter of interest then there are locally optimal unbiased estimating functions, but generally there is no globally optimal estimating function. We consider conditioning on the minimal sufficient statistic for the nuisance parameter and find the conditional linear optimal unbiased estimating function. Since the nuisance parameter is totally eliminated in the conditional model there is no intrinsic problem in setting up conditional tests of significance and confidence intervals. A compromise between conditional and unconditional optimum estimating functions is suggested. The techniques are illustrated on three examples including the well known common means problem. The proposed hypothesis testing and confidence interval procedures work reasonably well for the examples considered.     

Birnbaum‐Saunders autoregressive conditional range model applied to stock index data

This article proposes a new approach to the conditional autoregressive range (CARR) model using the Birnbaum‐Saunders (BS) distribution. The model aims to develop volatility clustering, which incorporates extreme fluctuations, using a time‐varying evolution of the range process called the BSCARR model. Furthermore, diagnosis analysis tools for diagnosis analysis were developed to evaluate the goodness of fit, such as residual analysis, global influence measures based on Cook's distance, and local influence analysis. For illustrative purposes, three real financial market indices are analyzed. A comparison with classical CARR models was also carried out in these examples. The results indicated that the proposed model outperformed some existing models in the literature, especially a recent CARR model based on the gamma distribution even under the presence of atypical cases (observed values).     

On Properties of the (Φ, a)-Power Divergence Family with Applications in Goodness of Fit Tests

In this paper we unify the different measures of divergence by introducing a general class of measures of divergence, the (Φ, a) −power divergence family and investigate its main properties including the limiting property, the order preserving property, and the quadratic convergence. For the practical implications of the proposed class of measures, we examine its use in goodness of fit tests for multinomial populations. In particular, a test statistic for goodness of fit tests based on the proposed family of measures is investigated for small sample sizes and various multinomial distributions that include symmetric, skewed and equiprobable models. The proposed statistic appears to work well in all cases considered as opposed to other traditional tests including the traditional chi-squared Pearson’s test, which may work well in some but not all situations.     

Gliding motion of bacteria on power‐law slime

The present investigation deals with an undulating surface model for the motility of bacteria gliding on a layer of non‐Newtonian slime. The slime being the viscoelastic material is considered as a power‐law fluid. A hydrodynamical model of motility involving an undulating cell surface which transmits stresses through a layer of exuded slime to the substratum is examined. The non‐linear differential equation resulting from the balance of momentum and mass is solved numerically by a finite difference method with an iteration technique. The manner in which the various exponent values of the power‐law flow affect the structure of the boundary layer is delineated. A comparison is made of the power‐law fluid with the Newtonian fluid. For the power‐law fluid with respect to different power‐law exponent values, shear‐thinning and shear‐thickening effects can be observed, respectively. Copyright © 2004 John Wiley & Sons, Ltd.     

A quadraticity limit theorem useful in linear models

Summary This paper proves an asymptotic uniform quadraticity of certain statistics based on randomly weighted residual empirical processes under fairly general conditions. This result is useful in many statistical inference problems pertaining to linear models including the Correlation and Autoregression models. Some applications to the goodness of fit tests and minimum distance estimation in linear models are given.     

Non‐linear Control of a Hydraulic Piston Actuator without Velocity Information

The mathematical models of hydraulic actuators are known to be non‐linear. Therefore, in order to increase the dynamic performance of the closed‐loop system, we have to take into account the significant non‐linearities of the hydraulic plant in the controller design. In this contribution, we deal with so‐called valve‐controlled translational piston actuators. In general, they have the pleasing property to be exact input‐to‐state linearizable in the sense of the differential geometric control synthesis approach. However, in practical applications it often turns out that those controllers, which have to rely on the knowledge of the piston velocity, have problems in the case of noisy measurements. This is why, we propose an approach where the non‐linear controller is designed in such a way that the control law is independent of the piston velocity. It can be even proven that the closed‐loop system is globally asymptotically stable.     

Conservation laws for Lotka–Volterra models

We derive necessary and sufficient conditions on a Lotka–Volterra model to admit a conservation law of Volterra's type. The result and the proof for the corresponding linear algebra problem are given in graph‐theoretical terms; they refer to the directed graph which is defined by the coefficients of the differential equation system. Copyright © 2003 John Wiley & Sons, Ltd.     

The log-likelihood ratio for sparse multinomial mixtures

The log likelihood ratio is expanded for testing a sequence of multinomial null hypotheses against a sequence of multinomial mixture close alternative hypothesis. As the number of categories grows without limit, the sample size increases and the variances of the mixing distributions tend to zero. The limiting form of the log likelihood ratio is functionally different from previously studied goodness of fit statistics. The statistic derived here exhibits moderate asymptotic power when Pearson's chi-square is biased.     

Global stability of a periodic Holling‐Tanner predator‐prey model

This paper is concerned with the global dynamics of a Holling‐Tanner predator‐prey model with periodic coefficients. We establish sufficient conditions for the existence of a positive solution and its global asymptotic stability. The stability conditions are first given in average form and afterward as pointwise estimates. In the autonomous case, the previous criteria lead to a known result.     

Goodness-of-Fit and Sufficiency: Exact and Approximate Tests

A procedure to test fit to a distribution where a minimal sufficient statistic is available, is discussed for testing the Poisson distribution. The test is exact, and is compared with a simpler approximate test. A remarkable correlation between the p-values given by the exact and the approximate procedures is found, and shows the power of the computer over and above what is usually acknowledged.     

Log‐symmetric regression models: information criteria and application to movie business and industry data with economic implications

This work deals with log‐symmetric regression models, which are particularly useful when the response variable is continuous, strictly positive, and following an asymmetric distribution, with the possibility of modeling atypical observations by means of robust estimation. In these regression models, the distribution of the random errors is a member of the log‐symmetric family, which is composed by the log‐contaminated‐normal, log‐hyperbolic, log‐normal, log‐power‐exponential, log‐slash and log‐Student‐t distributions, among others. One way to select the best family member in log‐symmetric regression models is using information criteria. In this paper, we formulate log‐symmetric regression models and conduct a Monte Carlo simulation study to investigate the accuracy of popular information criteria, as Akaike, Bayesian, and Hannan‐Quinn, and their respective corrected versions to choose adequate log‐symmetric regressions models. As a business application, a movie data set assembled by authors is analyzed to compare and obtain the best possible log‐symmetric regression model for box offices. The results provide relevant information for model selection criteria in log‐symmetric regressions and for the movie industry. Economic implications of our study are discussed after the numerical illustrations.     

Testing for Uniformity of the Residual Life Time Based on Dynamic Kullback-Leibler Information

In this article a goodness of fit test for distributional assumptions regarding the residual lifetime is proposed. The test is based on a Vasicek type sum log-spacings estimators of a dynamic version of Kullback-Leibler information. The specific distributional hypothesis considered is of the uniformity over [0,1]. However, the test can be used for testing any simple goodness of fit hypothesis. The asymptotic distribution of the test statistic together with a tabulation of the critical points for different sample sizes are given. Finally, the power function of the test is empirically studied in comparison with some competitors, and the test appears to be meritorious.     

Rank analogues of the likelihood ratio test for an ordered alternative in a two-way layout

Summary Distribution-free tests for no treatment effect against the simple order alternative in a two-way layout with equal number of observations per cell are considered. The nonparametric test statistics are constructed by the rank analogues of the likelihood ratio test statistic assuming normality (i) based on within-block rankings and (ii) based on combined rankings of all the observations after alignment within each block. The exact distributions are given and large sample properties are investigated. The asymptotic power of the test (i) as the number of observations per eell tends to infinity can be satisfied enough, and in the case that the number of blocks tends to infinity, the asymptotic power of the test (ii) is almost higher than that of the test (i). Also these rank tests are compared with linear rank tests and it is shown that these proposed tests are robust by a table.     

Supervision System of Machining Process States

Due to non‐linear, multidimensional and random character of most processes of machining, an use of analytical methods to process monitoring is difficult, time‐consuming and expensive. An application of algorithms which base on models in real processes is very limited, especially in the case of grinding processes. On the one hand, such methods require realistic and exact models of monitored process, on the other hand, they can characterize restrictive hypothesis concerning the process modeled. Most of model‐basing learning algorithms have an application to linear and steady‐state processes. However, majority of monitored processes are non‐linear, and, additionally, of nonstationary character. The system of monitoring proposed in the paper bases on artificial intelligence methods, which makes it possible to exclude from it a model of the grinding process.     

On Least Squares Estimation for Stable Nonlinear AR Processes

Following a Markov chain approach, this paper establishes asymptotic properties of the least squares estimator in nonlinear autoregressive (NAR) models. Based on conditions ensuring the stability of the model and allowing the use of a strong law of large number for a wide class of functions, our approach improves some known results on strong consistency and asymptotic normality of the estimator. The exact convergence rate is established by a law of the iterated logarithm. Based on this law and a generalized Akaike's information criterion, we build a strongly consistent procedure for selection of NAR models. Detailed results are given for familiar nonlinear AR models like exponential AR models, threshold models or multilayer feedforward perceptions.