[R] log-minus-log plot

[Marc Schwartz]

on 02/20/2009 07:23 AM Frank E Harrell Jr wrote:
> C.H. wrote:
>> Dear experts,
>>
>> I would like to know how to plot the log-minus-log plot for survival
>> analysis (to check the proportional assumption) in R.
>>
>> Using the AML example.
>>
>> fit <- survfit(Surv(time, status) ~ x, data=aml)
>> length(fit$surv) #20
>>
>> as the length of fit$surv is shorter than aml$x and aml$time. I don't
>> know how to plot.
>>
>> Thank you.
>>
>> Regards,
>>
>> CH
>>
>>
> 
> library(Design)  # uses library(survival)
> fit <- survfit(. . .)
> survplot(fit, loglog=TRUE)
> ?survplot for details


Just to throw out another possibility, using a Cox model on the data:

  library(survival)

  fit <- coxph(Surv(time, status) ~ x, data = aml)


  > cox.zph(fit)
                    rho   chisq     p
  xNonmaintained 0.0198 0.00691 0.934


  # create plot. default transform = "km"
  plot(cox.zph(fit, transform = "log"))


See ?coxph, ?cox.zph and ?plot.cox.zph in 'survival'.

HTH,

Marc Schwartz