Editorial to special issue Frontiers of Data Analysis

The special issue Frontiers of Data Analysis of the Journal of Applied Statistics (JAS), Taylor & Francis, contains papers that were presented in the 2019 ICSA China Conference which took place during July 1–4, 2019 on the campus of Nankai University, Tianjin, China. The conference was organized jointly by the International Chinese Statistical Association, Nankai University and Shanghai Jiaotong University. The conference was also co-sponsored by the K.C. Wong Education Foundation of Hong Kong and Chern Institute of Mathematics of Nankai University. The scientific program committee of the conference focused on creating collaboration opportunities and identifying new directions for further research. The conference attracted more than 500 participants and offered two plenary keynote lectures, 102 invited scientific sessions, and social events that included the opening mixer and banquet. The scientific sessions covered a wide range of topics including biostatistics, bioinformatics, statistics, engineering, finance, economics, genetics and genomics, big data computing, clinical trials, health policy and data science. This special issue of JAS includes 10 papers that were presented in the conference and were carefully reviewed by referees under the guideline of the journal. The selected papers cover various topics of state of art statistical methods and applications.

He, Sun and Shao [1] considered a failure time data analysis problem under the context of gene identification and presented a network-based survival analysis approach for identifying target genes that can be used for developing cancer immunotherapies and predicting patient survival. In the method, the Cox model was used and the LASSO estimation procedure was applied for estimation. Then the authors applied the approach for identifying candidate genes as possible targets for new intervention of gliomas based on both animal and human data.

Sub-cohort sampling designs including nested case-control (NCC) and case-cohort (CC) studies are common in biomedical studies. Lee, Zeleniuch-Jacquotte and Liu [2] presented Monte Carlo simulation studies on risk prediction performance under NCC, CC and full-cohort studies to illustrate the importance of the matching procedure in these designs.

In statistical process control (SPC), the quality of a process can be characterized by a functional relationship between two or more variables, particularly by a profile monitoring. Liu, Zhu and Lin [3] studied a generalized likelihood ratio test for monitoring a profile data through a nonparametric regression by estimating the on-line profiles without a prespecified functional form for the profiles.

Wang, Jia and Jin [4] discussed regression analysis of bivariate right-censored failure time data or univariate right-censored failure time data with dependent censoring with the focus on estimation of the cumulative baseline hazard function. More specifically, they investigated the situation where the failure time of interest follows the Cox model and the relationship between two correlated failure times or the failure time of interest and the censoring time can be described by an Archimedean copula model. In addition to providing a useful formula for the estimation, the authors also developed a graphical model checking procedure.

Wang et al. [5] investigated the inference about the class of Box-Cox transformation models, which is commonly used for regression analysis of failure time data due to its flexibility, when one observes left-truncated and right-censored failure time data. In particular, the proportional hazards model and the additive hazards model are two special cases. The authors proposed a Bayesian estimation approach for the situation where the baseline hazards function can be approximated by the piecewise function. In the method, a conditional marginal prior was employed and a MCMC sampling procedure was developed.

Yang et al. [6] addressed the multiplicity issue in the multiple comparisons in correlated binary data. They present asymptotic simultaneous confidence intervals (SCIs) for many-to-one comparisons of proportion differences adjusting for multiplicity and the correlation, which can be applied in the analysis of paired data arising in ophthalmological, orthopedic and otolaryngologic studies.

The Cox proportional hazards regression model has been used extensively in social science, biomedical sciences, economics and other fields. Zhao and Su [7] extended the Cox proportional hazards regression model with a generalized concept of relative risk and presented a novel application on how to identify factors associated with the housing price. With the housing data in Tianjin, a metropolitan city in China, they presented the results that how the macro regulation policies, qualities of public schools, the structure and neighborhood characteristics and distance of the residential property to downtown Tianjin were associated with the housing price.

The order-of-addition experiment arises in many areas including biochemistry, pharmaceutical science and food and nutritional science. It is to identify the optimal order of adding components for optimizing the response of interest. With large m components, it is not feasible to test all m! possible adding order. Zhao et al. [8] provided a study on pair-wise ordering designs and presented a new class of minimal-point pair-wise ordering designs using a recursive relation between two successive full pair-wise ordering designs.

Interval-censored failure time data are a general type of failure time data and include right-censored failure time data as a special case. Correspondingly a great deal of literature has been established for regression analysis of interval-censored data. In contrast, only limited literature exists on regression analysis of mixed interval-censored, the topic of Zhu et al. [9], a combination of exact and interval-censored observations. For the problem, the authors investigated the fitting of the proportional odds model to the data and derived the maximum likelihood estimation procedure. Furthermore, they established the asymptotic properties of the proposed estimators and applied the method a motivating set of data on child cancer survivors.

As failure time data, longitudinal data also often occur in various studies and in various forms. Yu and Zhong [10] discussed regression analysis of high-dimensional longitudinal data under the time-varying linear mixed effect model and presented a penalized estimation procedure with the use of both lasso penalty and fusion penalty. In particular, an efficient two-stage parameter estimation algorithm was provided for the estimation of longitudinal trajectories of fixed effects coefficients, and the approach was applied to a set of health and retirement survey data.

Acknowledgments

We would like to thank the authors and reviewers for their time and effort. We are indebted to the Editor-in-Chief, Dr. Jie Chen, and her editorial team for their editorial assistance.