Feature extraction in mixture cure model with broken adaptive ridge
The mixture cure model (MCM) is used in the presence of a cure fraction in identifying features associated with a time-to-event outcome. In the field of biomedical research, high-dimensional survival datasets are common and hence feature extraction is key to various scientific discoveries. However,...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/166472 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The mixture cure model (MCM) is used in the presence of a cure fraction in identifying features associated with a time-to-event outcome. In the field of biomedical research, high-dimensional survival datasets are common and hence feature extraction is key to various scientific discoveries. However, there exist few variable selection methods currently for MCMs under high-dimensional settings where there are more predictors than samples. This study proposes a dual iterative algorithm, the expectation-maximization – broken adaptive ridge (EM-BAR), for high-dimensional penalized Weibull MCM in identifying factors associated with cure status and survival. In comparison to popular regularization methods such as LASSO and ridge, BAR is asymptotically consistent for variable selection, possesses an oracle property for parameter estimation in a sparse model, and acquires a grouping effect for highly correlated variables. Various signal strengths were considered. Through extensive simulation studies, the penalized MCM has been shown to identify a high proportion of true signals (high power) for prognostic factors associated with both cure status and survival time. |
|---|