R/kmplt.R

In RCASPAR: A package for survival time prediction based on a piecewise baseline hazard Cox regression model.

#### Plotting a Kaplan Meier plot
#### Input is expeceted to be a table with at least the domains Patient, survival.time, censored (censorship status), title=title for the generated graph
#### Censorship status can be either False or True.
kmplt<-function(data, title) {
    no.subjects <- nrow(data)
    new.row <- rep(NA,ncol(data))
    data.new <- rbind(new.row, data)
    data.new$True_STs[1] <- 0
    data.new$censored[1] <- TRUE
    data.new <- data.new[order(data.new$True_STs), ]
    data.new$censored[data.new$censored == TRUE] <- 1
    #### ri=patients at risk and mi=patients who die
    risk <- c(0,no.subjects:1)
    deaths <- c()
    pt<-c(NA) #### proportion of patinets surviving at time t
    St<-c(1) #### cumulitave proporiton of patients surviving at time t;S(t)
    for(death in 1:(no.subjects+1)){
        if(data.new$censored[death] == 1) 
           mi <- 0
        else
             mi <- 1
    deaths <- c(deaths, mi)
    }
    data.new <- cbind(data.new, risk, deaths)
    for(patient in 2:(no.subjects + 1)) {
        pi <- (risk[patient]-deaths[patient])/risk[patient]
        St <- c(St, St[patient - 1]*pi)
        pt <- c(pt, pi)
        }
    data.new <- cbind(data.new, pt, St)
    y_axis <- c(data.new$St[length(data.new$True_STs)])
    x_axis <- data.new$True_STs
    for(survival in c(no.subjects:1)) {
        if(data.new$True_STs[survival] != data.new$True_STs[survival+1]) 
            y_axis <- c(y_axis, data.new$St[survival])
        else y_axis <- c(y_axis, y_axis[length(y_axis)])
    }
    y_axis <- y_axis[length(y_axis):1]
    return(plot(x_axis, y_axis, type="S", xlab="Time", ylab="Probability of survival S(t)", main=title))
}