Some Exploratory Tools for Survival Analysis

Abstract

Graphical and computer-intensive methods for exploring survival data are becoming more widely available, though perhaps not yet widely used. We review some recent developments in this area, including running quantile plots, local estimation of the relative risk function, and recursive partitioning. The use of Martingale residuals in some of these methods will be contrasted with other approaches. The methods are applied to data on patients with multiple myeloma treated on trials conducted by the Southwest Oncology Group.