Survival analysis methods are widely used in medical research: for example in clinical trials, in pharmacological studies, in studying the effects of ionizing radiation, in screening for breast and colon rectal cancer, and in the study of sexually transmitted diseases. Recently there has been considerable interest in the biostatistics literature regarding the use of splines to estimate the log-hazard function. We propose to incorporate recent and ongoing advances in this area into HARE, a polynomial spline survival analysis module to be integrated into S-Plus, a modern statistical analysis system. When the observed failure times are generated by some unknown mechanism, interest often lies in estimating a hazard, survival, or density function. In HARE these functions will be estimated by means of polynomial splines and their selected tensor products in univariate and bivariate survival models with covariates. HARE will also compute standard and not--so-- standard test statistics. The graphical display of the resulting estimates and diagnostics will be comprehensive. An algorithm employing stepwise addition and deletion of basis functions makes adaptive survival function estimation possible in HARE. Spline estimation can be (and will be) applied to other models such as multiple logistic regression and classification.
Survival is one of the most heavily used statistical methods. The HARE module adds an exciting new dimension to survival analysis. StatSci intends to target the biomedical and reliability engineering markets with HARE. Each of these markets represents tens of thousands of potential users. It is difficult to envision a commercial situation in the statistical software arena that is more likely to succeed.
