• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.109-128
• /
• 2020

A climate-impact projection usually consists of several stages, and the uncertainty of the projection is known to be quite large. It is necessary to assess how much each stage contributed to the uncertainty. We call an uncertainty quantification method in which relative contribution of each stage can be evaluated as uncertainty decomposition. We propose a new Bayesian model for uncertainty decomposition in climate change impact assessments. The proposed Bayesian model can incorporate uncertainty of natural variability and utilize data in control period. We provide a simple and efficient Gibbs sampling algorithm using the auxiliary variable technique. We compare the proposed method with other existing uncertainty decomposition methods by analyzing streamflow data for Yongdam Dam basin located at Geum River in South Korea.

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.1-14
• /
• 2020

Investigations of biosimilarity between reference drugs and test drugs required statistical tests; in addition, statistical tests to evaluate biosimilarity have been recently proposed. Ordinal outcome data has been observed in research; however, appropriate statistical tests to deal with ordinal endpoints for biosimilar have not yet been proposed. This paper extends existing design for ordinal endpoints. Using measure of nominal-ordinal association and relative distances between drugs are defined so that testing procedures are developed. Through simulation studies, we investigate type I error rate and power to show the performance of our suggested method. Furthermore, a comparison between the statistical tests and other designs is proviede to show significance of ordinal endpoints.

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.129-140
• /
• 2020

In this paper, we introduce an efficient algorithm for the non-convex penalized multinomial logistic regression that can be uniformly applied to a class of non-convex penalties. The class includes most non-convex penalties such as the smoothly clipped absolute deviation, minimax concave and bridge penalties. The algorithm is developed based on the concave-convex procedure and modified local quadratic approximation algorithm. However, usual quadratic approximation may slow down computational speed since the dimension of the Hessian matrix depends on the number of categories of the output variable. For this issue, we use a uniform bound of the Hessian matrix in the quadratic approximation. The algorithm is available from the R package ncpen developed by the authors. Numerical studies via simulations and real data sets are provided for illustration.

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.37-46
• /
• 2020

Recurrent event data frequently occur in clinical studies, demography, engineering reliability and so on (Cook and Lawless, The Statistical Analysis of Recurrent Events, Springer, 2007). Sometimes, two or more different but related type of recurrent events may occur simultaneously. In this study, our interest is to estimate the covariate effect on bivariate recurrent event times with zero inflations. Such zero inflation can be related with susceptibility. In the context of bivariate recurrent event data, furthermore, such susceptibilities may be different according to the type of event. We propose a joint model including both two intensity functions and two cure rate functions. Bivariate frailty effects are adopted to model the correlation between recurrent events. Parameter estimates are obtained by maximizing the likelihood derived under a piecewise constant hazard assumption. According to simulation results, the proposed method brings unbiased estimates while the model ignoring cure rate models gives underestimated covariate effects and overestimated variance estimates. We apply the proposed method to a set of bivariate recurrent infection data in a study of child patients with leukemia.

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.15-35
• /
• 2020

In research on behavioral studies, significant attention has been paid to the stage-sequential process for multiple latent class variables. We now explore the stage-sequential process of multiple latent class variables using the multivariate latent class profile analysis (MLCPA). A latent profile variable, representing the stage-sequential process in MLCPA, is formed by a set of repeatedly measured categorical response variables. This paper proposes the extended MLCPA in order to explain an association between the latent profile variable and the latent group variable as a form of a two-dimensional contingency table. We applied the extended MLCPA to the National Longitudinal Survey on Youth 1997 (NLSY97) data to investigate the association between of developmental progression of depression and substance use behaviors among adolescents who experienced Authoritarian parental styles in their youth.

• Communications for Statistical Applications and Methods
• /
• 제26권6호
• /
• pp.527-538
• /
• 2019

This paper considers the issue of obtaining the optimal design in Poisson regression model when the sample size is small. Poisson regression model is widely used for the analysis of count data. Asymptotic theory provides the basis for making inference on the parameters in this model. However, for small size experiments, asymptotic approximations, such as unbiasedness, may not be valid. Therefore, first, we employ the second order expansion of the bias of the maximum likelihood estimator (MLE) and derive the mean square error (MSE) of MLE to measure the quality of an estimator. We then define DM-optimality criterion, which is based on a function of the MSE. This criterion is applied to obtain locally optimal designs for small size experiments. The effect of sample size on the obtained designs are shown. We also obtain locally DM-optimal designs for some special cases of the model.

• Communications for Statistical Applications and Methods
• /
• 제26권6호
• /
• pp.583-589
• /
• 2019

Counterexamples of a skew-normal distribution are developed to improve our understanding of this distribution. Two examples on bivariate non-skew-normal distribution owning marginal skew-normal distributions are first provided. Sum of dependent skew-normal and normal variables does not follow a skew-normal distribution. Continuous bivariate density with discontinuous marginal density also exists in skew-normal distribution. An example presents that the range of possible correlations for bivariate skew-normal distribution is constrained in a relatively small set. For unified skew-normal variables, an example about converging in law are discussed. Convergence in distribution is involved in two separate examples for skew-normal variables. The point estimation problem, which is not a counterexample, is provided because of its importance in understanding the skew-normal distribution. These materials are useful for undergraduate and/or graduate teaching courses.

• Communications for Statistical Applications and Methods
• /
• 제26권6호
• /
• pp.539-556
• /
• 2019

This paper derive a method to solve change point regression problems via a process for obtaining consequential results using properties of a difference-based intercept estimator first introduced by Park and Kim (Communications in Statistics - Theory Methods, 2019) for outlier detection in multiple linear regression models. We describe the statistical properties of the difference-based regression model in a piecewise simple linear regression model and then propose an efficient algorithm for change point detection. We illustrate the merits of our proposed method in the light of comparison with several existing methods under simulation studies and real data analysis. This methodology is quite valuable, "no matter what regression lines" and "no matter what the number of change points".

• Communications for Statistical Applications and Methods
• /
• 제27권1호
• /
• pp.47-64
• /
• 2020

In this paper, we derive some estimators of the scale parameter of the exponentiated half-logistic distribution based on the multiply Type-I hybrid censoring scheme. We assume that the shape parameter λ is known. We obtain the maximum likelihood estimator of the scale parameter σ. The scale parameter is estimated by approximating the given likelihood function using two different Taylor series expansions since the likelihood equation is not explicitly solved. We also obtain Bayes estimators using prior distribution. To obtain the Bayes estimators, we use the squared error loss function and general entropy loss function (shape parameter q = -0.5, 1.0). We also derive interval estimation such as the asymptotic confidence interval, the credible interval, and the highest posterior density interval. Finally, we compare the proposed estimators in the sense of the mean squared error through Monte Carlo simulation. The average length of 95% intervals and the corresponding coverage probability are also obtained.

• Communications for Statistical Applications and Methods
• /
• 제26권6호
• /
• pp.591-610
• /
• 2019

Deep Learning is one of the machine learning methods to find features from a huge data using non-linear transformation. It is now commonly used for supervised learning in many fields. In particular, Convolutional Neural Network (CNN) is the best technique for the image classification since 2012. For users who consider deep learning models for real-world applications, Keras is a popular API for neural networks written in Python and also can be used in R. We try examine the parameter estimation procedures of Deep Neural Network and structures of CNN models from basics to advanced techniques. We also try to figure out some crucial steps in CNN that can improve image classification performance in the CIFAR10 dataset using Keras. We found that several stacks of convolutional layers and batch normalization could improve prediction performance. We also compared image classification performances with other machine learning methods, including K-Nearest Neighbors (K-NN), Random Forest, and XGBoost, in both MNIST and CIFAR10 dataset.