The semiparametric accelerated failure time (AFT) model is a log-linear model of failure times with an unspecified random error term. The rank-based estimator has been a popular estimation method for regression parameters in this model. An induced smoothing method has reduced computational complexity and instability in the original non-smooth rank-based estimator. This paper briefly reviews and compares the recently proposed computationally efficient variance estimation methods for the semiparametric AFT models in multivariate failure times settings. Comparisons are made via extensive simulation experiments. Based on our findings, we may recommend using ‘Diff-Boot’ and ‘Diff-Closed’ methods with a one-step iteration. These variance estimators are then illustrated with the well-known Diabetic retinopathy study data.
|Number of pages||24|
|Journal||Japanese Journal of Statistics and Data Science|
|Publication status||Published - 2021 Dec|
Bibliographical noteFunding Information:
This work was partly supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2020R1A2C1A0101313911) and the Graduate School of YONSEI University Research Scholarship Grants in 2020.
© 2021, Japanese Federation of Statistical Science Associations.
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- Computational Theory and Mathematics