On the morning of May 11, at the invitation of the School of Statistics, Professor Xuewen Lu from the University of Calgary gave an online lecture entitled Sieve estimation of a class of partially linear transformation models with interval-censored competing risks data for teachers and students of the School of Statistics . This lecture was chaired by Liu Xiaohui, Deputy Dean of the School of Statistics, and some teachers and graduate students of the School of Statistics attended the lecture.

At the beginning of the lecture, Professor Lu Xuewen first explained what is competitive risk data through specific examples. Competing risk data means that in biomedical studies with time-event outcomes, each subject can experience any of several individual events, whichever occurs first.Its biggest feature is that the time point of the event cannot be accurately observed, and only the time point of its occurrence can be known within a certain period of time. Prof. Lu then uses a dataset of HIV-infected individuals in sub-Saharan Africa to specify competing risk data.

Then, Professor Lu Xuewen introduced a class of partial linear transformation models with interval-censored competing risk data, which was mainly studied in his thesis. Optimal estimates of a large number of parametric and nonparametric components in the model are obtained by maximizing the likelihood function on the sieve space of a joint B-spline and Bernstein polynomial under the semi-parametric generalized odds ratio setting of the cause-specific cumulative incidence function. This setting considers a relatively simple finite-dimensional parameter space and an infinite-dimensional parameter space that can be approximated by a finite-dimensional parameter space. When the sample size tends to infinity, it can be proved that the estimates of all parameters converge almost everywhere. Asymptotic normal distribution and semiparametric validity of estimates of velocity, and finite-dimensional parameters. Finally, Professor Lu explained the limited sample nature of the method and its practical value through simulation experiments and empirical data.

At the end of the lecture, Professor Lu Xuewen mentioned that there are still many challenges in using the partial linear transformation model to deal with the competing risk data at the end of the interval in real life. For example, in practical applications, the connection function is usually unknown, and how to choose a suitable one of the problems solved. Second, how to determine which covariates are parametric and which are nonparametric in a partially linear GOR transform model, that is, how to determine g, which represents the number of unknown smooth regression functions. To solve this problem, Lu Jiashou proposed two solutions, one of which is to simply put discrete covariates in the linear part and continuous covariates in the nonlinear part. The second is the screening method, that is, a preliminary univariate analysis is performed first, and then continuous covariates are separated according to the shape of the estimated nonparametric function. It is easy to know that the screening method is more reasonable.

Professor Lu Xuewen's wonderful speech this time fully demonstrated the theoretical and practical analysis of a class of partial linear transformation models with interval-censored competitive risk data, providing a useful example for teachers and students of the School of Statistics to engage in related research work. (Photo/Peng Yuqi Text/Dai Jingyi)

[Extended reading]

The keynote speaker, Professor Lu Xuewen, is currently a professor of statistics in the Department of Mathematical Statistics at the University of Calgary, Canada, and a doctoral tutor.He received a bachelor's degree in mathematics from Hunan Normal University in 1987, a master's degree in probability and statistics from Peking University in 1990, and a doctorate degree from Guelph University in Canada in 1997.Work experience: 1990-1994 as a lecturer at the Department of Mathematics at Hunan Normal University, 1997-2002 as a biostatistician at the Canadian Federal Ministry of Agriculture and Agri-Food, 2008.08-2009.06 as a visiting associate professor at the University of Michigan Department of Biostatistics, 2002.7-2013.03 in Calgary University as an assistant professor and associate professor, 2013.04 to present, as a tenured professor at the University of Calgary.His research interests include nonparametric/semiparametric regression, censored regression, survival analysis, biomedical statistics, generalized linear/additive models, mixed models, panel data analysis, empirical likelihood methods, variable selection, and machine learning.So far, dozens of high-level academic papers have been published in Technometrics, Scandinavian Journal of Statistics, Statistica Sinica, Journal of Multivariate Analysis, Computational Statistics and Data Analysis, Journal of Cerebral Blood Flow and Metabolism, International Journal of Food Microbiology, etc.Participated in the editing of two monographs: Modeling Microbial Responses in Food>, 2003, CRC Press and Advanced Statistical Methods in Data Science, 2016, Springer. He is also an associate editor of the Journal of Statistical Computation and Simulation.