1

I tried to get confident interval of hazard ratio in survreg() function, but it's not that straightforward as coxph. 

Call:
survreg(formula = Surv(survival, DIED) ~ AGE + GENDER + PLATE + 
NEUTRO + NIH, data = IMRAWandIST, dist = "exponential")
           Value Std. Error     z         p
PLATE        0.00236   0.000367  6.42  1.39e-10
EUTRO      -0.02726   0.016695 -1.63  1.03e-01

Scale fixed at 1

Exponential distribution Loglik(model)= -4628.6 Loglik(intercept only)= -4736.1 Chisq= 215 on 5 degrees of freedom, p= 0 Number of Newton-Raphson Iterations: 5 n=917 (28 observations deleted due to missingness)

#estimate of beta
a <- c(coefficients(summary(fitexp)))

print(coef <- (a * -1 * 1 / 1)

#estimate of HR

print(HR <- exp(coef))

The result doesn't have CI, only standard deviation, so my question is how can I transform SE of AFT coefficient into PH coefficient then compute the CI of HR.

I'm kinda stuck here. Can someone help?

Konrad Rudolph
  • 530,221
  • 131
  • 937
  • 1,214
Nicole Wang
  • 11
  • 2
  • 2
    Some data to reproduce the example will be helpful – A. Suliman Apr 23 '18 at 05:29
  • It works for me. library(survival); fitexp<- survreg(Surv(futime, fustat) ~ ecog.ps + rx, ovarian, dist='weibull',scale=1);a<-c(coefficients(summary(fitexp)));print(coef<-(a*-1*1/1));print(HR<-exp(coef)) – A. Suliman Apr 23 '18 at 05:38
  • The coefficients and std errors are in the "table"- component of the list returned by `summary`. It's pretty much exactly like the case with `coxph`. I have no idea what the expression `(a * -1 * 1 / 1)` is supposed to represent. – IRTFM Apr 23 '18 at 07:21
  • Call: survreg(formula = Surv(survival, DIED) ~ AGE + GENDER + PLATE + NEUTRO + NIH, data = IMRAWandIST, dist = "exponential") Value Std. Error z p (Intercept) 9.61783 0.284733 33.78 4.09e-250 AGE -0.03443 0.003855 -8.93 4.16e-19 Scale fixed at 1 Exponential distribution Loglik(model)= -4628.6 Loglik(intercept only)= -4736.1 Chisq= 215 on 5 degrees of freedom, p= 0 Number of Newton-Raphson Iterations: 5 n=917 (28 observations deleted due to missingness) – Nicole Wang Apr 23 '18 at 14:08
  • the result looks something like this, only has standard deviation, not CI. – Nicole Wang Apr 23 '18 at 14:09

0 Answers0