Hello,   I am trying to estimate a sample size for the log rank test comparing two different treatment arms. The study will have a three year followup when the effect of the treatment will change after each year. the input I am using is:   proc power;    twosamplesurvival test=logrank       Alpha = 0.05       curve("trt1") = 1:0.65 2:0.53625 3:0.489328125         curve("trt2") = (1 to 3 by 1):(0.82 0.7462 0.712621)       groupsurvival = "trt1" | "trt2"       accrualtime = 2       Totaltime = 3         groupmedlosstimes = 10 | 10         groupweights = (1 2)         power = 0.90       EVENTSTOTAL=. ; Run;   This program designs a piecewise linear effect. Can I choose an ascending or descending linear effect? Also, I want to define a piecewise exponential? And last, Can I define the effect in hazard ratio terms and not survival curve but with a changing hazard ratio over time?   Thank you! ... View more