TY - JOUR
T1 - Estimating the occurrence rate for prevalent survival data in competing risks models
AU - Huang, Ying
AU - Wang, Mei Cheng
N1 - Funding Information:
Ying Huang is Assistant Biostatistician, Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, NY 10021. Mei-Cheng Wang is Associate Professor, Department of Biostatistics, Johns Hopkins University, Baltimore, MD 21205. The second author’s research is partially supported by National Institutes of Health Grants R01-A129197 and R01-AI33744. The authors thank the associate editor and a referee for the comments that greatly improved the presentation.
Copyright:
Copyright 2015 Elsevier B.V., All rights reserved.
PY - 1995/12
Y1 - 1995/12
N2 - The problem of interest is the estimation of occurrence probabilities based on prevalent data in competing risks models. In the literature, the development of nonparametric methods has relied heavily on the independent competing risks assumption. The primary purpose of this article is to establish a statistical framework without excluding the possibility of dependence among competing risks. This is done through the use of crude hazard functions. The crude hazard functions not only are estimable regardless of whether the competing risks are independent, but also are mathematically more tractable. In this article we show that there is a one-to-one correspondence between the crude hazard functions and the occurrence probabilities. The general inversion formulas for the occurrence probabilities are presented, from which various representations can be derived under different sampling techniques. Maximum likelihood estimators are derived using these representations for nonparametric and length bias data. The maximum likelihood property and asymptotic behavior of both estimation procedures are studied. The simulation results show that the length bias estimators have smaller variances compared to the nonparametric estimators. Nevertheless, the nonparametric estimation procedure appears to be more robust to model assumptions.
AB - The problem of interest is the estimation of occurrence probabilities based on prevalent data in competing risks models. In the literature, the development of nonparametric methods has relied heavily on the independent competing risks assumption. The primary purpose of this article is to establish a statistical framework without excluding the possibility of dependence among competing risks. This is done through the use of crude hazard functions. The crude hazard functions not only are estimable regardless of whether the competing risks are independent, but also are mathematically more tractable. In this article we show that there is a one-to-one correspondence between the crude hazard functions and the occurrence probabilities. The general inversion formulas for the occurrence probabilities are presented, from which various representations can be derived under different sampling techniques. Maximum likelihood estimators are derived using these representations for nonparametric and length bias data. The maximum likelihood property and asymptotic behavior of both estimation procedures are studied. The simulation results show that the length bias estimators have smaller variances compared to the nonparametric estimators. Nevertheless, the nonparametric estimation procedure appears to be more robust to model assumptions.
KW - Cross-sectional sample
KW - Crude hazard functions
KW - Length bias data
KW - Maximum likelihood estimation
KW - Occurrence probabilities
UR - http://www.scopus.com/inward/record.url?scp=21344445440&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=21344445440&partnerID=8YFLogxK
U2 - 10.1080/01621459.1995.10476646
DO - 10.1080/01621459.1995.10476646
M3 - Article
AN - SCOPUS:21344445440
SN - 0162-1459
VL - 90
SP - 1406
EP - 1415
JO - Journal of the American Statistical Association
JF - Journal of the American Statistical Association
IS - 432
ER -