R/predictCoxPL.R

In bozenne/riskRegressionLight: Light version of the package riskRegression

### predictCoxPL.R --- 
#----------------------------------------------------------------------
## author: Brice Ozenne
## created: sep  4 2017 (16:43) 
## Version: 
## last-updated: aug 28 2018 (10:42) 
##           By: Brice Ozenne
##     Update #: 151
#----------------------------------------------------------------------
## 
### Commentary: 
## 
### Change Log:
#----------------------------------------------------------------------
## 
### Code:

## * predictCoxPL (documentation)
#' @title Computation of survival probabilities from Cox regression models using the product limit estimator.
#' @description Same as predictCox except that the survival is estimated using the product limit estimator.
#' @name predictCoxPL
#'
#' @inheritParams predictCox
#' @param ... additional arguments to be passed to \code{\link{predictCox}}.
#' 

## * predictCoxPL (code)
#' @rdname predictCoxPL
#' @examples 
#' library(survival)
#'
#' #### generate data ####
#' set.seed(10)
#' d <- sampleData(40,outcome="survival")
#' nd <- sampleData(4,outcome="survival")
#' d$time <- round(d$time,1)
#'
#' #### Cox model ####
#' fit <- coxph(Surv(time,event)~ X1 + X2 + X6,
#'              data=d, ties="breslow", x = TRUE, y = TRUE)
#'
#' ## exponential approximation
#' predictCox(fit, newdata = d, times = 1:5)
#' 
#' ## product limit
#' predictCoxPL(fit, newdata = d, times = 1:5)
#'
#' #### stratified Cox model ####
#' fitS <- coxph(Surv(time,event)~ X1 + strata(X2) + X6,
#'              data=d, ties="breslow", x = TRUE, y = TRUE)
#'
#' ## exponential approximation
#' predictCox(fitS, newdata = d, times = 1:5)
#' 
#' ## product limit
#' predictCoxPL(fitS, newdata = d, times = 1:5)
#' 
#' #### fully stratified Cox model ####
#' fitS <- coxph(Surv(time,event)~ 1,
#'              data=d, ties="breslow", x = TRUE, y = TRUE)
#'
#' ## product limit
#' GS <- survfit(Surv(time,event)~1, data = d)
#' range(predictCoxPL(fitS)$survival - GS$surv)
#'
#' fitS <- coxph(Surv(time,event)~ strata(X2),
#'              data=d, ties="breslow", x = TRUE, y = TRUE)
#'
#' ## product limit
#' GS <- survfit(Surv(time,event)~X2, data = d)
#' range(predictCoxPL(fitS)$survival - GS$surv)
#'

## * predictCoxPL (code)
#' @rdname predictCoxPL 
#' @export
predictCoxPL <- function(object,
                         times,
                         newdata = NULL, 
                         type = c("cumhazard","survival"),
                         keep.strata = TRUE,
                         keep.infoVar = FALSE,
                         ...){
    stop <- NULL ## [:CRANcheck:] data.table
    
    ## ** normalize arguments
    object.modelFrame <- coxModelFrame(object)
    if(!is.null(newdata)){
        if(is.data.table(newdata)){
            newdata <- copy(newdata)
        }else{
            newdata <- as.data.table(newdata)
        }
    }else if (missing(times)) {
        times <- numeric(0)
    }
    
    if("survival" %in% type == FALSE){
        stop("The argument \'type\' must contain \"survival\" \n",
             "use the function predictCox otherwise \n")
    }

    ## ** run original prediction
    original.res <- predictCox(object = object,
                               newdata = newdata,
                               times = times,
                               type = type,
                               keep.strata = TRUE,
                               keep.infoVar = TRUE,
                               ...)
    infoVar <- original.res$infoVar
    if(keep.infoVar==FALSE){
        original.res$infoVar <- NULL
    }

    ## ** compute survival
    if(is.null(newdata)){
        hazard.res <- as.data.table(predictCox(object = object,
                                               type = "hazard",
                                               keep.strata = TRUE,
                                               ...))

        if(!is.null(hazard.res$strata)){
            Ustrata <- levels(hazard.res$strata)
            n.Ustrata <- length(Ustrata)
            
            for(iS in 1:n.Ustrata){ ## iS <- 1
                iIndex.all <- which(hazard.res$strata==Ustrata[iS])
                iIndex.times <- which(original.res$strata==Ustrata[iS])
            
                vec.surv <- c(1,cumprod(1-hazard.res$hazard[iIndex.all]))
           
                index.jump <- prodlim::sindex(eval.times = original.res$times[iIndex.times],
                                              jump.times = c(0,hazard.res$times[iIndex.all]))
            
                original.res$survival[iIndex.times] <- vec.surv[index.jump]
            }
            if(keep.strata == FALSE){
                original.res$strata <- NULL
            }
        }else{
            vec.surv <- c(1,cumprod(1-hazard.res$hazard))
           
            index.jump <- prodlim::sindex(eval.times = original.res$times,
                                          jump.times = c(0,hazard.res$times))
            
            original.res$survival <- vec.surv[index.jump]
        }
        
        if("hazard" %in% type == FALSE){
            original.res$hazard <- NULL
        }
        
    }else if(infoVar$is.strata){
        object.strata <- coxStrata(object,
                                   data = NULL,
                                   strata.vars = infoVar$stratavars)
        new.strata <- original.res$strata
        if(keep.strata == FALSE){
            original.res$strata <- NULL
        }
        Ustrata <- unique(new.strata)
        n.Ustrata <- length(Ustrata)

        for(iStrata in 1:n.Ustrata){ # iStrata <- 1
            indexStrata.object <- which(object.strata==Ustrata[iStrata])
            indexStrata.newdata <- which(new.strata==Ustrata[iStrata])
                
            all.times <- object.modelFrame[indexStrata.object,.SD$stop]
            all.times <- sort(unique(all.times[all.times <= max(times)]))

            if(length(all.times)>0){
                res.tempo <- predictCox(object,
                                        newdata = newdata[indexStrata.newdata,],
                                        times = all.times,
                                        type = "hazard")

                lastEventTime.tempo <- res.tempo$lastEventTime[which(infoVar$strata.levels==Ustrata[iStrata])]
                
                if(0 %in% all.times){                    
                    hazard.tempo <- cbind(res.tempo$hazard,NA)
                    all.timesA <- c(all.times,lastEventTime.tempo + 1e-10)
                }else{
                    hazard.tempo <- cbind(0,res.tempo$hazard,NA)
                    all.timesA <- c(0,all.times,lastEventTime.tempo + 1e-10)
                }

                if(original.res$diag){
                    index.jump <- prodlim::sindex(eval.times = times[indexStrata.newdata], jump.times = all.timesA)
                    original.res$survival[indexStrata.newdata,] <- cbind(sapply(1:length(indexStrata.newdata), function(iP){ # iP <- 16
                        prod(1-hazard.tempo[iP,1:index.jump[iP]])
                    }))
                }else{
                    index.jump <- prodlim::sindex(eval.times = times, jump.times = all.timesA)
                    survival.PL <- t(apply(hazard.tempo, 1, function(x){cumprod(1-x)}))
                    original.res$survival[indexStrata.newdata,] <-  survival.PL[,index.jump,drop=FALSE]
                }
            }
        }
        
    }else{
        all.times <- unique(sort(object.modelFrame[stop <= max(times), stop]))
        if(length(all.times)>0){
            res.tempo <- predictCox(object,
                                    newdata = newdata,
                                    times = all.times,
                                    type = "hazard")
            if(0 %in% all.times){
                hazard.tempo <- cbind(res.tempo$hazard,NA)
                all.timesA <- c(all.times,res.tempo$lastEventTime+1e-10)
            }else{
                hazard.tempo <- cbind(0,res.tempo$hazard,NA)
                all.timesA <- c(0,all.times,res.tempo$lastEventTime+1e-10)
            }

            index.jump <- prodlim::sindex(eval.times = times, jump.times = all.timesA)
            if(original.res$diag){
                original.res$survival <- cbind(sapply(1:NROW(newdata), function(iP){ # iP <- 16
                    prod(1-hazard.tempo[iP,1:index.jump[iP]])
                }))
            }else{
                survival.PL <- t(apply(hazard.tempo, 1, function(x){cumprod(1-x)}))
                original.res$survival <-  survival.PL[,index.jump,drop=FALSE]
            }
        }
    }

    ## ** export
    return(original.res)
    
    
}
#----------------------------------------------------------------------
### predictCoxPL.R ends here