I wanted to determine functional form of some of my variables for Cox analysis and I was playing with the stan2 database. The Martingale residuals indicate that log(wait) is a better form. But when I did the Cox-Snell residual vs Nelson-Aalen cumulative hazard plots with both wait and log(wait), wait seems to be a better fit.
Based on this, how do I determine whether to transform something like wait? I was following along with the Cleves, Gould & Marchenko Survival book chapter 11 and hence my question in the forum.
Based on this, how do I determine whether to transform something like wait? I was following along with the Cleves, Gould & Marchenko Survival book chapter 11 and hence my question in the forum.
