Congratulations to Yi Yu who has been awarded a UOB- CUHK exchange programme grant.
Survival analysis is an important branch of statistics studying the time-to-event data. For instance, in a clinical trial, one would like to know if a specific treatment will significantly increase life expectancy of patients of a certain disease. The main difficulty here is the existence of censoring, i.e., the clinical trial most likely terminates before all the patients die. This phenomenon by no means only occurs in clinical trials. For example, in finance, the deaths of the patients can be replaced by the bankruptcies of companies. In the classic setting, the number of covariates is small, but with the development of technology, researchers have access to huge amounts of different covariates; therefore, the need to develop new statistical tools, and theoretical justifications thereof, is in high demand.
In our current collaborat project, we are interested in directed and dynamic communication networks, with information and disease spreading as our motivations. We model the interaction in each possible pair using recurrent event data analysis, which is an important survival analysis tool, and globally use composite likelihood methods to take the network effect into consideration. Compared to the few existing papers tackling similar problems, instead of adding neighbourhood information in an ad hoc fashion, we use the composite likelihood to capture the network structure. In the sense of few model assumptions, our method is robust against model mis-specification.
Based on this collaboration, our proposed research is motivated by the convictions that 1) a statistically rigorous framework of high-dimensional survival analysis is within reach, and 2) due to the increasing demand from the application areas, our proposed research will have a high impact on both the U.K. and the Hong Kong. The work proposed will improve the understanding of existing methods and facilitate the development of novel techniques, and it revolves around three themes: 1) outlier detection in the high-dimensional survival analysis models, 2) change point detection in high-dimensional survival models, and 3) high-dimensional quantile regression problems in survival models.
To the best of our knowledge, despite the close relationship between some parts of the UoB and the CUHK, Dr Sit and I are the only connection between the Mathematics and/or Statistics areas of the two universities. We would like to strengthen this established connection, and we believe visiting each other will also help to make a broader connection at the Department and/or School level.
For more information on the grant see here.
