We will develop flexible survival models which capture crossed survival curves and spatiotemporal information. By flexibly integrating spatiotemporal factors with possible risks, the proposed models will improve the prediction of cancer risk in practice. The generalized accelerated failure time (GAFT) model can be used when the standard proportional hazards model and the accelerated failure time model cannot be used, which common problem in spatial survival is modeling. Furthermore, we will incorporate spatiotemporal information into the GAFT model, which can improve the prediction of cancer survival varying by region. All proposed models will be estimated through semiparametric Bayesian estimation methods, enabling the proposed model to be a useful alternative to geoadditive model based on other semiparametric survival model within standard classes such as PH or AFT.
The specific aims are :
Aim 1 : Develop GAFTST models and associated semiparametric estimation methods based on the linear dependent tailfree process and evaluate the use of the proposed models by simulation study. ? Aim 2: Develop user-friendly software within the R environment that will allow researchers and practitioners to apply the advanced modeling strategies directly to available cancer data sets. ? Aim 3: We will examine the effects of time, geographical region and racial disparities of prostate cancer data set of Louisiana from Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute (NCI). The statistical models developed are innovative because they can reveal the complex relationship among longitudinal and spatial patterns, risk factors and prostate cancer mortality based on the easily interpretable models and findings. Moreover, the software development will facilitate the use of the proposed advanced statistical tools by other researchers to study the relationship among spatiotemporal patterns, risk factors and other forms of cancer. Therefore, the proposed research will markedly improve the understanding of the effects between longitudinal spatial patterns and cancer mortality, and thus expand the body of knowledge on how to implement cost-effective screening, treat and counsel individuals more effectively, and plan community-wide intervention strategies.
In this project, we propose to investigate the association between longitudinal spatial patterns, risk factors, and prostate cancer mortality though development of advanced, flexible survival models that allow for crossing survival curves and spatiotemporal information. Moreover, the software development in R will facilitate the use of the proposed advanced statistical tools by other researchers to study the relationship among longitudinal spatial patterns, risk factors, and other forms of cancer. Therefore, the proposed research will markedly improve the understanding of the effects between longitudinal spatial patterns and cancer mortality, and thus expand the body of knowledge on how to implement cost-effective screening, treat and counsel individuals more effectively, and plan the community-wide intervention strategies.
| Zhang, Jiajia; Hanson, Timothy; Zhou, Haiming (2018) Bayes factors for choosing among six common survival models. Lifetime Data Anal : |
| Zhou, Haiming; Hanson, Timothy; Zhang, Jiajia (2017) Generalized accelerated failure time spatial frailty model for arbitrarily censored data. Lifetime Data Anal 23:495-515 |
| Li, Li; Hanson, Timothy; Zhang, Jiajia (2015) Spatial extended hazard model with application to prostate cancer survival. Biometrics 71:313-22 |
| Zhou, Haiming; Hanson, Timothy; Jara, Alejandro et al. (2015) MODELLING COUNTY LEVEL BREAST CANCER SURVIVAL DATA USING A COVARIATE-ADJUSTED FRAILTY PROPORTIONAL HAZARDS MODEL. Ann Appl Stat 9:43-68 |
