||5R01CA061937-06 Interpret this number
||University Of N Carolina At Chapel Hill
||Logspline Methods for Survival Analysis
This proposal requests funding for the development of new statistical
methodologies for research in survival analysis.
Survival time data arise in many clinical and medical studies of
diseases. One of the primary variables is the time to some event such
as recurrence of the disease or death. The time to event or survival
time is right-censored if the event has not occurred before the end of
the study period. Statistical analyses for these studies are required
to include covariates that may vary with time. The study subjects may
constitute a selection-biased or random truncation sample. Also, the
survival times may be correlated as in the study of occurrences of
multiple diseases or in studies where subjects are family members. The
primary purpose of this proposal is to address these issues by
developing methods of statistical inference based on extended linear
modeling, and to do so in a manner that effectively balances the bias
and variance of the estimates. The proposal has seven projects. The
first project will conduct a theoretical investigation of the bootstrap
method based on-the nonadaptive approach to the saturated HARE models.
The second project considers the problem of estimating covariate effect
function in proportional hazards modeling. The third project establish
distributional properties and standard errors of estimates of the
transition intensities for event history analysis. The fourth project
develops an extended linear model in estimating the density.
conditional density, regression and hazard regression functions for
length-biased data. The fifth project considers a new approach to
randomly truncated data. Methods of extended linear models are used to
model the underlying density and conditional density functions. The
sixth project considers a new approach to longitudinal data analysis
using extended linear models. The seventh project considers a new
approach to multivariate survival data analysis using extended linear
models. The strengths and weaknesses of the proposed procedures will
be critically examined by simulations and theoretical investigations.
Software will be developed to analyze clinical data from cancer and
cardiovascular studies. Distributional properties will be established
and standard errors of the estimates of effects will be provided and
will be used to lend support to the software development.
The origin of DNA replication and Fieller's problem.
, Scott R.S.
, Vos J.M.
Statistics in medicine, 2002-12-15; 21(23), p. 3571-82.
Nonparametric regression estimates with censored data: local linear smoothers and their applications.
, Truong Y.K.
Biometrics, 1998 Dec; 54(4), p. 1434-44.
Hazard regression with interval-censored data.
, Clarkson D.B.
Biometrics, 1997 Dec; 53(4), p. 1485-94.