[R] Building a web risk calculator based on Cox PH--definitive method for calculating probability?

vokey588 jahan.mohiuddin at gmail.com
Tue Jul 17 16:13:29 CEST 2012


Hello all,
I am a medical student and as a capstone for my summer research project I am
going to create a simple online web "calculator" for users to input their
relevant data, and a probability of relapse within 5 years will be computed
and returned based on the Cox PH model I have developed.

The issue I'm having is finding a definitive method/function to feed the
user's "newdata" and return the probability of relapse within 5 years.  I
have googled this and the answers seems to be inconsistent; I have variously
seen people recommend survest(), survfit(), and predict.coxph().  Terry had
a useful answer 
http://r.789695.n4.nabble.com/how-to-calculate-predicted-probability-of-Cox-model-td4515265.html
here  but I didn't quite understand what he meant in his last sentence.  

Here is some code for you to quickly illustrate what you suggest.  

library(rms)
library(survival)
library(Hmisc)
data(veteran)
dd=datadist(veteran)
options(datadist='dd')
options(digits=4)
obj=with(veteran,Surv(time,status))
vetcoxph=coxph(obj~celltype+trt,data=veteran)    #I will fit models from
both the survival and rms packages so you can 
#use what you like
vetcph=cph(obj~celltype+trt,data=veteran,surv=TRUE,time.inc=5*365,x=T,y=T)
#let's say the user inputted that their cell type was smallcell and their
treatment was "1".
userinput=data.frame(celltype='smallcell',trt=factor(1))

I really appreciate your recommendations

Best,
Jahan




--
View this message in context: http://r.789695.n4.nabble.com/Building-a-web-risk-calculator-based-on-Cox-PH-definitive-method-for-calculating-probability-tp4636749.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list