| fixcoxph {PHInfiniteEstimates} | R Documentation |
Remove observations from a proportional hazards regression, and return the fit of the reduced model.
Description
This function implements the approximate conditional inferential approach of Kolassa and Zhang (2019) to proportional hazards regression.
Usage
fixcoxph(randdat, xxx, iv, verbose = FALSE)
Arguments
randdat |
A list with at least the component y, representing the Surv() object. I expect that this will be output from an initial non-convergent regression. |
xxx |
a design matrix for the regression. I expect that this will be the $x component of the output from an initial non-convergent regression, run with x=TRUE . |
iv |
name of the variable of interest, as a character string |
verbose |
logical flag governing printing. |
Value
Fitted survival analysis regression parameter of class coxph, fitted form data set with observations forcing infinite estimation removed.
References
Kolassa JE, Zhang J (2019). https://higherlogicdownload.s3.amazonaws.com/AMSTAT/fa4dd52c-8429-41d0-abdf-0011047bfa19/UploadedImages/NCB_Conference/Presentations/2019/kolassa_toxslides.pdf. Accessed: 2019-07-14.
Examples
data(breast) # From library coxphf
bcfit<-coxph(Surv(TIME,CENS)~ T+ N+ G+ CD,data=breast,x=TRUE)
fixcoxph(bcfit,bcfit$x,"T",Surv(TIME,CENS)~ T+ N+ G+ CD)
testdat2 <- data.frame(Time=c(4,3,1,1,2,2,3),
Cen=c(1,1,1,0,0,0,0), Primary=c(0,2,1,1,1,0,0), Sex=c(0,0,0,0,1,1,1))
(bcfit<-coxph(Surv(Time,Cen)~Primary + Sex, testdat2, x=TRUE, ties="breslow"))
fixcoxph(bcfit,bcfit$x,"Primary")