Create new columns in loop based on different datasets

Question

Here's a brief description of the data I have: Survival data from 4 separate studies that compares the survival rates among 20 groups. Each study lasted a different amount of time. For example, study 1 lasted 42 days and Study 2 lasted 50 days.

Here's a snapshot of the data:

UniqueID    Time    Censored    Group1     Group2      Study
ABC123      6       1           1          111         1
DEF456      42      0           1          112         1
GHI789      42      0           2          344         1 
JKL012      38      1           2          564         1
MNO345      19      1           10         761         1
PQR678      13      1           5          222         2
STU901      5       1           20         333         2
VWX234      50      0           15         444         2 
YZA567      20      1           15         555         2
BCD890      50      0           12         555         2

Here's what I want to do: I want to create a function that allows the user to select two parameters (Study, Group1) to compare survival rates.

This is what I have attempted so far:

SurvA=function(a,b){
   setwd("path to my file")
   data=read.xlsx("mydata.xlsx",sheet=1)
   data_study$Study==a
   list(unique(data_study$Group1))
}

I want to write a loop that scans the list for all the unique Group1 numbers and create Group1 specific variables with the following logic as an example:

data_study$Group1_10=ifelse(data_study$Group1==10,1,0)
data_study$Group1_12=ifelse(data_study$Group1==12,1,0)

I'm unsure of how to proceed with the loop that would make this happen.

Once that is finalized, the rest of the code would look like this:

 library(survival)
 library(survminer)
 SurvA=function(a,b){
   setwd("path to my file")
   data=read.xlsx("mydata.xlsx",sheet=1)
   data_study$Study==a
   list(unique(data_study$Group1))

   #LOOP

   surv_object=Surv(time=data_study$Time,event=data_study$Censored)
   fit=survfit(surv_object~b,data=data_study)
   ggsurv=ggsurvplot(fit,data=data_study,pval=TRUE,xlim=c(0,60),
              title='Study 'a' Survival Plot for Group 'b' ',xlab="Time (days)")
   ggsurv$plot=ggsurv$plot+theme(plot.title=element_text(hjust=0.5))
   print(ggsurv)

}

Any help would be appreciated! Also, if you have suggestions for more efficient ways to write what I've already got - I would be very happy to learn of better ways to do this.

Answer 1

Ultimately it sounds like you are operating on a data frame that represents the results of a bunch of studies. You want to write a function that takes as input a study identifier and a patient group within that study, and you want the function to plot the survival curve for patients in the specified group versus not.

Since your function only needs to handle a single specified group b, it seems simplest to me to just create a single variable indicating membership in that group or not, instead of looping through all variables as you propose:

library(survival)
library(survminer)
SurvA <- function(dat, a, b) {
  dat <- dat[dat$Study == a,]
  dat$Group1Val <- ifelse(dat$Group1 == b, b, paste("Not", b))
  fit <- survfit(Surv(Time, Censored)~Group1Val, data=dat)
  print(ggsurvplot(fit, data=dat, pval=TRUE,
                   title=paste("Study", a, "Survival Plot for Group", b),
                   xlab="Time (Days)",
                   ggtheme=theme(plot.title=element_text(hjust=0.5))))
}
SurvA(dat, 1, 1)

Result:

Data:

set.seed(144)
s1g1S <- rexp(100, 1) ; s1g1C <- rexp(100, 0.5) ; s1g2S <- rexp(100, 1.2) ; s1g2C <- rexp(100, 0.5)
s2g1S <- rexp(100, 1) ; s2g1C <- rexp(100, 0.5) ; s2g2S <- rexp(100, 1.2) ; s2g2C <- rexp(100, 0.5)
dat <- data.frame(UniqueID=seq_len(200),
                  Time=c(pmin(s1g1S, s1g1C), pmin(s1g2S, s1g2C), pmin(s2g1S, s2g1C), pmin(s2g2S, s2g2C)),
                  Censored=as.numeric(c(s1g1S, s1g2S, s2g1S, s2g2S) > c(s1g1C, s1g2C, s2g1C, s2g2C)),
                  Group1=rep(c(1, 2, 1, 2), each=100), Study=rep(1:2, each=200))