# What to do if my -log(-log(S(t|x))) kaplan-meier survival function violates PH assumption?

I've already performed survdiff() on my survival curves to find that there is a difference in time-to-germination between my 3 treatments. As far as I'm aware, I will need to run coxph() to be able to determine how different my treatments are from eachother. However, my survival function violates PH assumption (at least I think so).

This is what I have done:

attach(data)

mlogmlog<-function(y){-log(-log(y))}

GR<-survfit(Surv(Days[which(Site=="2")],Status[which(Site=="2")])~Treat[which(Site=="2")],type="kaplan-meier")

plot(GR, mark.time=F, fun=mlogmlog, log="x", xlab="t[days]",ylab="-log(-log(S(t)))", col=c("blue","red","green"), lwd=1.75) It looks like the red treatment is crossing.

When clear violations of the PH assumption are detected, various remedies are available in standard statistical software that may resolve the problem and still permit use of the Cox model. These include converting covariates with non-proportional effects into stratiﬁcation factors or time-dependent covariates, and dividing the time axis into discrete segments and analysing data for one or more of the segments separately (see Therneau and Grambsch, 2000).

this was taken from McNeal et al, however I am not entirely sure what this means, and I've looked up Therneau and Grambsch, 2000, but the authors use a different stats program from R.

Any help will be much appreciated! Thanks.

• What are the curves in the plot? The cumulative hazards are given by -log(S(t)), and perhaps that is an easier way to assess proportionality. – Theodor Jun 24 '16 at 21:42