This could be a naive question; however, I am just a beginner in doing survival analysis. I stratified patient samples into tertile (t1, t2, and t3) based on a known oncogene's expression values. I wanted to draw kaplan-meier (KM) survival curves for t1, t2, and t3. Thereafter, I wanted to compute log-rank test p-value between t2 and t1, and between t3 and t1. My question is when I compute log-rank test p-value between t2 and t1, should I consider only those samples that belong to t2 and t1? Should I proceed with a separate piece of analysis after excluding samples that belong to t3 group? Seeking your advice.