Kwara State University
Kazeem Dauda
Designation: Lecturer I
Department: Statistics and Mathematical Sciences
| My Publications | ||||||||
|---|---|---|---|---|---|---|---|---|
| S/N | Title | Abstract | Authors | Volume Numbers | Publication Type | Publication Date | Link | |
| 1 | Novel Decision Tree-Based Survival Analysis with Simultaneous Competing Risks |
The development of joint competing risks in tree-based survival analysis is not much been employed by the various researchers. The physician sometimes may want to know the behavior of patients with similar risk via both events of interest and the competing event simultaneously. Previously, many researchers have considered one event at a time out of the kth competing events and adopted some splitting rules on both within-node homogeneity and between-node heterogeneity. However, these techniques do not allow simultaneous comparison of the risk group, which sometimes may be of interest to clinicians. In this study, marginal distribution function (MF), conditional incident function (CIF) and Log-Odds Residuals for cumulative incident function (RF) were proposed and used in the development of tree-based survival analysis with competing risks. In particular, the analytical relationship between these methods was provided via the sum-of-square split functions, and this tentatively explains the broadly observed similarity of the resulting trees. Additional evaluation was carried out using already published real-life data on liver transplant and Hodgkin’s disease. Comparisons of these techniques to existing methods were provided using the two real-life data. It is observed that the new methods improve upon the existing classical method and survival tree-based with competing risks were well developed. | Dauda, K. A. | Technical Report No. - MIU/TR-02/2018 | Technical Report | 2018-03-12 | ||