Computational Statistics最新文献

Bayesian Kriging (BK) provides a way to estimate regression models where the parameters are smoothed across space. Such estimates could help guide site-specific fertilizer recommendations. One advantage of BK is that it can readily fill in the missing values that are common in yield monitor data. The problem is that previous methods are too computationally intensive to be commercially feasible when estimating a nonlinear production function. This paper sought to increase computational speed by imposing restrictions on the spatial covariance matrix. Previous research used an exponential function for the spatial covariance matrix. The two alternatives considered are the conditional autoregressive and simultaneous autoregressive models. In addition, a new analytical solution is provided for finding the optimal value of nitrogen with a stochastic linear plateau model. A comparison among models in the accuracy and computational burden shows that the restrictions significantly reduced the computational burden, although they did sacrifice some accuracy in the dataset considered.

This paper studies a Bayesian local influence method to detect influential observations in a partially linear model with first-order autoregressive skew-normal errors. This method appears suitable for small or moderate-sized data sets ((n=200{sim }400)) and overcomes some theoretical limitations, bridging the diagnostic gap for small or moderate-sized data in classical methods. The MCMC algorithm is employed for parameter estimation, and Bayesian local influence analysis is made using three perturbation schemes (priors, variances, and data) and three measurement scales (Bayes factor, (phi )-divergence, and posterior mean). Simulation studies are conducted to validate the reliability of the diagnostics. Finally, a practical application uses data on the 1976 Los Angeles ozone concentration to further demonstrate the effectiveness of the diagnostics.

Quantile regression is an extension of linear regression which estimates a conditional quantile of interest. In this paper, we propose an empirical likelihood-based non-parametric procedure to detect structural changes in the quantile regression models. Further, we have modified the proposed smoothed empirical likelihood-based method using adjusted smoothed empirical likelihood and transformed smoothed empirical likelihood techniques. We have shown that under the null hypothesis, the limiting distribution of the smoothed empirical likelihood ratio test statistic is identical to that of the classical parametric likelihood. Simulations are conducted to investigate the finite sample properties of the proposed methods. Finally, to demonstrate the effectiveness of the proposed method, it is applied to urinary Glycosaminoglycans (GAGs) data to detect structural changes.

Additive coefficient models generalize linear regression models by assuming that the relationship between the response and some covariates is linear, while their regression coefficients are additive functions. Because of its advantages in dealing with the “curse of dimensionality”, additive coefficient models gain a lot of attention. The commonly used estimation methods for additive coefficient models are not robust against high leverage points. To circumvent this difficulty, we develop a robust variable selection procedure based on the exponential squared loss function and group penalty for the additive coefficient models, which can tackle outliers in the response and covariates simultaneously. Under some regularity conditions, we show that the oracle estimator is a local solution of the proposed method. Furthermore, we apply the local linear approximation and minorization-maximization algorithm for the implementation of the proposed estimator. Meanwhile, we propose a data-driven procedure to select the tuning parameters. Simulation studies and an application to a plasma beta-carotene level data set illustrate that the proposed method can offer more reliable results than other existing methods in contamination schemes.

This paper proposes a polynomial structure identification (PSI) method for variable selection and model structure identification of additive models with longitudinal data. First, the backfitting algorithm and zero-order local polynomial smoothing method are used to select important variables in the additive model, and the importance of variables is determined through the inverse of the bandwidth parameter in the nonparametric partial kernel function. Second, the backfitting algorithm and Q-order local polynomial smoothing method are utilized to identify the specific structure of each selected predictor. To incorporate correlations within longitudinal data, a two-stage estimation method is proposed for estimating the regression parameters of the identified important variables: (i) Parameter estimators of the important variables are firstly obtained under an independence working model assumption; (ii) Generalized estimating equations with a working correlation matrix based on B-splines are constructed to obtain the final estimators of the parameters, which improve the efficiency of parameter estimation. Finally, simulation studies are carried out to evaluate the performance of the proposed method, followed by the presentation of two real-world examples for illustration.

In time-to-event analyses, a competing risk is an event whose occurrence precludes the occurrence of the event of interest. Settings with competing risks occur frequently in clinical research. Missing data, which is a common problem in research, occurs when the value of a variable is recorded for some, but not all, records in the dataset. Multiple Imputation (MI) is a popular method to address the presence of missing data. MI uses an imputation model to generate M (M > 1) values for each variable that is missing, resulting in the creation of M complete datasets. A popular algorithm for imputing missing data is multivariate imputation using chained equations (MICE). We used a complex simulation design with covariates and missing data patterns reflective of patients hospitalized with acute myocardial infarction (AMI) to compare three strategies for imputing missing predictor variables when the analysis model is a cause-specific hazard when there were three different event types. We compared two MICE-based strategies that differed according to which cause-specific cumulative hazard functions were included in the imputation models (the three cause-specific cumulative hazard functions vs. only the cause-specific cumulative hazard function for the primary outcome) with the use of the substantive model compatible fully conditional specification (SMCFCS) algorithm. While no strategy had consistently superior performance compared to the other strategies, SMCFCS may be the preferred strategy. We illustrated the application of the strategies using a case study of patients hospitalized with AMI.

In this work, we present a methodology to estimate the strength of handball teams. We propose the use of the Conway-Maxwell-Poisson distribution to model the number of goals scored by a team as a flexible discrete distribution which can handle situations of non equi-dispersion. From its parameters, we derive a mathematical formula to determine the strength of a team. We propose a ranking based on the estimated strengths to compare teams across different championships. Applied to female handball club data from European competitions over the 2022/2023 season, we show that our new proposed ranking can have an echo in real sports events and is linked to recent results from European competitions.

Hypothesis testing for the regression coefficient associated with a dichotomized continuous covariate in a Cox proportional hazards model has been considered in clinical research. Although most existing testing methods do not allow covariates, except for a dichotomized continuous covariate, they have generally been applied. Through an analytic bias analysis and a numerical study, we show that the current practice is not free from an inflated type I error and a loss of power. To overcome this limitation, we develop a bootstrap-based test that allows additional covariates and dichotomizes two-dimensional covariates into a binary variable. In addition, we develop an efficient algorithm to speed up the calculation of the proposed test statistic. Our numerical study demonstrates that the proposed bootstrap-based test maintains the type I error well at the nominal level and exhibits higher power than other methods, as well as that the proposed efficient algorithm reduces computational costs.

This article focuses on the practical issue of a recent theoretical method proposed for trend estimation in high dimensional time series. This method falls within the scope of the low-rank matrix factorization methods in which the temporal structure is taken into account. It consists of minimizing a penalized criterion, theoretically efficient but which depends on two constants to be chosen in practice. We propose a two-step strategy to solve this question based on two different known heuristics. The performance and a comparison of the strategies are studied through an important simulation study in various scenarios. In order to make the estimation method with the best strategy available to the community, we implemented the method in an R package TrendTM which is presented and used here. Finally, we give a geometric interpretation of the results by linking it to PCA and use the results to solve a high-dimensional curve clustering problem. The package is available on CRAN.

In this paper we discuss nonlinear dimensionality reduction within the framework of principal curves. We formulate dimensionality reduction as problems of estimating principal subspaces for both noiseless and noisy cases, and propose the corresponding iterative algorithms that modify existing principal curve algorithms. An R squared criterion is introduced to estimate the dimension of the principal subspace. In addition, we present new regression and density estimation strategies based on our dimensionality reduction algorithms. Theoretical analyses and numerical experiments show the effectiveness of the proposed methods.