Degree Name

BS

Department

Statistics

College

Physical and Mathematical Sciences

Defense Date

2023-04-26

Publication Date

2023-06-05

First Faculty Advisor

Natalie Jean Blades

First Faculty Reader

Gregory Snow

Honors Coordinator

Del Scott

Keywords

left truncation, rank tests, survival analysis

Abstract

In survival analysis, there is a potential interaction between left-truncated data and rank test procedures for the difference between two groups that has not yet been investigated in depth. This study explores the effect of left truncation on the absolute and relative performance of the unweighted log-rank test and the weighted Prentice modification of the Gehan-Wilcoxon test. Resulting power from simulation study replications revealed that while left truncation does not affect which test performs better overall, it does affect the relative difference between the tests. It significantly worsens the individual performance of the Prentice-modified test, likely because of left truncation on the survival curve by which the test is weighted. Left truncation was also associated with an unusual decrease in power with increasing sample size for the log-rank test under non-proportional hazards

Download

Share

COinS