Journal of the Korean Statistical Society最新文献

This study aims to improve the predictive performance for the event time through the machine learning model and find informative variables in the time-to-event data, simultaneously. To address this issue, after regarding the time-to-event data as the dichotomized counting processes data for predicting survival time, we consider the time-dependent support vector machine (SVM) framework for the dichotomized counting process data, where the decision function in this framework consists of the time-independent risk score and time-dependent intercept. Also, we consider the empirical partial derivative of the risk score function with respect to each marginal predictor as the indicator for the important predictor. Through this approach, it is possible to predict survival time and find variables that affect on the survival time at the same time. Simulation studies were conducted to confirm the performance of the model, and real data analysis was conducted by predicting the survival time of the lung cancer after the diagnosis and selecting genes associate with lung cancer through human gene data.

This study examined the test of independence of survival time from a covariate in a more general setting using empirical process techniques. Previous research has been extended in several ways: (1) allow incompleteness of observation owing to censoring (2) allow the time-dependent covariate (3) allow the non-uniform covariate (4) prove the validity of weighted bootstrap to implement the proposed testing procedure. Certain classes of test statistics that are functionals of a natural empirical process were studied, and the limiting distribution of these statistics was then derived using the functional delta method. The limiting distributions included some linear functionals of zero mean tight Brownian bridges under the null hypothesis, and the tests were consistent against general alternatives. Tests implemented using weighted bootstrap were shown to be valid. The proposals are illustrated via simulation studies and an application to acute leukemia data.

This paper considers the asymptotic properties of the error density estimator in the nonlinear autoregressive models with (alpha)-mixing errors. The asymptotic distribution and uniform convergence rate for the kernel density estimator of error density function are obtained. Last, some simulations of histograms, confidence intervals and mean integrated square errors are illustrated, which agree with our theoretical results.

We propose a confidential approach for disseminating frequency tables constructed for any combination of key variables in the given microdata, including those of hierarchical key variables. The system generates all possible frequency tables by either marginalizing or aggregating fully joint frequency tables of key variables while protecting the original cells with low frequencies through two masking steps: the small cell adjustments for joint tables followed by the proposed algorithm called information loss bounded aggregation for aggregated cells. The two-step approach is designed to control both disclosure risk and information loss by ensuring the k-anonymity of original cells with small frequencies while keeping the loss within a bounded limit.

Ridge calibration is a penalized method used in survey sampling to reduce the variability of the final set of weights by relaxing the linear restrictions. We proposed a method for selecting the penalty parameter that minimizes the estimated mean squared error of the mean estimator when estimated auxiliary information is used. We showed that the proposed estimator is asymptotically equivalent to the generalized regression estimator. A simple simulation study shows that our estimator has the smaller MSE compared to the traditional calibration ones. We applied our method to predict election result using National Barometer Survey and Korea Social Integration Survey.

It is well-known that the fused lasso signal approximator (FLSA) is inconsistent in change point detection under the presence of staircase blocks in true mean values. The existing studies focus on modifying the FLSA model to remedy this inconsistency. However, the inconsistency of the FLSA does not severely degrade the performance in change point detection if the FLSA can identify all true change points and the estimated change points set is sufficiently close to the true change points set. In this study, we investigate some asymptotic properties of the FLSA under the assumption of the noise level (sigma _n = o(n log n)). To be specific, we show that all the falsely segmented blocks are sub-blocks of true staircase blocks if the noise level is sufficiently low and a tuning parameter is chosen appropriately. In addition, each false change point of the optimal FLSA estimate can be associated with a vertex of a concave majorant or a convex minorant of a discrete Brownian bridge. Based on these results, we derive an asymptotic distribution of the number of false change points and provide numerical examples supporting the theoretical results.

In this paper, we investigate the maximum likelihood estimator (MLE) for the parameter (theta) in the probability density function (f(x;theta )). We specifically focus on the application of moving extremes ranked set sampling (MERSS) and analyze its properties in large samples. We establish the existence and uniqueness of the MLE for two common distributions when utilizing MERSS. Our theoretical analysis demonstrates that the MLE obtained through MERSS is, at the very least, as efficient as the MLE obtained through simple random sampling with an equivalent sample size. To substantiate these theoretical findings, we conduct numerical experiments. Furthermore, we explore the implications of imperfect ranking and provide a practical illustration by applying our approach to a real dataset.

Shape analysis is widely used in many application areas such as computer vision, medical and biological studies. One challenge to analyze the shape of an object in an image is its invariant property to shape-preserving transformations. To measure the distance or dissimilarity between two different shapes, we worked with the square-root velocity function (SRVF) representation and the elastic metric. Since shapes are inherently high-dimensional in a nonlinear space, we adopted a tangent space at the mean shape and a few principal components (PCs) on the linearized space. We proposed classification methods based on logistic regression using these PCs and tangent vectors with the elastic net penalty. We then compared its performance with other model-based methods for shape classification in application to shape of algae in watersheds as well as simulated data generated by the mixture of von Mises-Fisher distributions.

Decentralized federated learning based on fully normal nodes has drawn attention in modern statistical learning. However, due to data corruption, device malfunctioning, malicious attacks and some other unexpected behaviors, not all nodes can obey the estimation process and the existing decentralized federated learning methods may fail. An unknown number of abnormal nodes, called Byzantine nodes, arbitrarily deviate from their intended behaviors, send wrong messages to their neighbors and affect all honest nodes across the entire network through passing polluted messages. In this paper, we focus on decentralized federated learning in the presence of Byzantine attacks and then propose a unified Byzantine-resilient framework based on the network gradient descent and several robust aggregation rules. Theoretically, the convergence of the proposed algorithm is guaranteed under some weakly balanced conditions of network structure. The finite-sample performance is studied through simulations under different network topologies and various Byzantine attacks. An application to Communities and Crime Data is also presented.

Abstract

This study considers the online monitoring problem for detecting the parameter change in time series of counts. For this task, we construct a monitoring process based on the residuals obtained from integer-valued generalized autoregressive conditional heteroscedastic (INGARCH) models. We consider this problem within a more general framework using martingale difference sequences as the monitoring problem on GARCH-type processes based on the residuals or score vectors can be viewed as a special case of the monitoring problems on martingale differences. The limiting behavior of the stopping rule is investigated in this general set-up and is applied to the INGARCH processes. To assess the performance of our method, we conduct Monte Carlo simulations. A real data analysis is also provided for illustration. Our findings in this empirical study demonstrate the validity of the proposed monitoring process.