Assistance with coxph and standard errors

Question

[roba6848]

Hi,

I am running a cox regression with coxph in R The regression runs fine and reports standard errors on the coefficients too. This is good. But I need to run a number of simulation runs and collect the standard errors. The standard errors are printed when I print the object:but I can find no way of getting them from the coxph.object returned by coxph. There is no mention of them in the coxph.object documentation:

> coxph.result
Call:
coxph(formula = Surv(timeall, event = PatientDeceased) ~ treatment + 
    AgeAtIndexDate + BMI + cellcount)

                     coef  exp(coef)   se(coef)      z      p
treatmentA      5.057e-01  1.658e+00  2.153e-01  2.349 0.0188
treatmentB    8.014e-02  1.083e+00  2.465e-01  0.325 0.7451
AgeAtIndexDate  4.626e-05  1.000e+00  2.258e-05  2.048 0.0405
BMI       -5.023e-03  9.950e-01  1.939e-02 -0.259 0.7956
cellcount         2.113e-02  1.021e+00  9.299e-02  0.227 0.8202

Likelihood ratio test=8.88  on 5 df, p=0.114
n= 137, number of events= 123

Can anyone tell me where to get the standard errors?

[curious]

I am not sure how familiar with R you are and I don't have a computer in front of my to check this, but I think you should just be able to go:

result <- coxph.result

to store results in a variable, then you could go:

typeof(result)

to print to the console what type of object it is, then look up indexing/slicing that object on google/stackoverflow. you probably need to make a function to get the specific elements that you want.

EDIT: Sorry I forgot to leave this as a comment not an answer :(

Answer 1

[Gordon Smyth]

In statistics, the term "standard error" means an estimated standard deviation. A standard deviation is just the square root of a variance, and the variance matrix of the coefficients is documented in the coxph.object help page.

You can extract the standard errors from coxph.result by

SE <- sqrt( diag( coxph.result$var ))