How to detect survival groups?

Question

[MatthewP]

Hello, given TCGA data how can I separate patients into 2(high, low) or more groups with statistics method?
This paper has mention Interval Coded Score index(ICS) and Rpart
I want to know what methods usually used to detect survival risk group? And want some ready-to-use tools recommend. I already found RPART TOOL, but not ICS. Thanks.

[MatthewP]

Or maybe I just sort the survival data by time(died and censored together) then devide by median?

Answer 1

[Kevin Blighe]

Any method is used to divide the cohort into strata for high|mid|low, et cetera. That is, there is no standard that is used, and neither should there be due to the fact that this is exploratory type work.

You can try:

• quartiles
• tertiles
• binary classification (based on median or something else)
• Upper 10%, mid 80%, and lower 10% of the range, et cetera
• Z-score cut-offs

Kevin

[MatthewP]

My concern is treating died and censored data as the same if I split survival time simplified.