A new approach of subgroup identification for high-dimensional longitudinal data

Abstract

Discovering a medication that suitable for all patients is not possible due to the fact that the reaction to medication may differ significantly across different patient subgroups. The heterogeneity of treatment effects is central to the agenda for both personalized medicine and treatment selection. To expedite the development of tailored therapies and improve the treatment efficacy, identification of subgroups that exhibit different treatment effects is thus playing an essential role. In this paper, we consider high-dimensional dense longitudinal observations which have frequent and large number of measurements with high-dimensional covariates. We offer a data-driven subgroup identification method, which incorporates the sparse boosting algorithm into homogeneity pursuit via change point detection. Extensive simulations are carried out to examine the performance of our proposed approach. We further illustrate our method by analyzing a wallaby growth dataset.

Keywords:

• Change point detection
• homogeneity pursuit
• sparse boosting
• minimum descriptive length
• longitudinal data
• personalized medicine

Acknowledgements

The authors acknowledge supports from the National Natural Science Foundation of China 11601447 and 11771066; the China Scholarship Council 201707005036.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors acknowledge supports from the National Natural Science Foundation of China [grant numbers 11601447 and 11771066]; the China Scholarship Council [grant number 201707005036].