| Title: | Survival Trees to Fit Left-Truncated and Right-Censored and Interval-Censored Survival Data |
|---|---|
| Description: | Recursive partition algorithms designed for fitting survival tree with left-truncated and right censored (LTRC) data, as well as interval-censored data. The LTRC trees can also be used to fit survival tree with time-varying covariates. |
| Authors: | Wei Fu, Jeffrey Simonoff, Wenbo Jing |
| Maintainer: | Wenbo Jing <[email protected]> |
| License: | GPL-3 |
| Version: | 1.1.1 |
| Built: | 2024-10-31 06:52:47 UTC |
| Source: | CRAN |
.logrank_trafo transforms Surv(time1, time2, event) objects into
logrank scores, which will be used later in the tree algorithm. It is
not designed to be used by users, not for internal used of LTRCIT
function.
.logrank_trafo2(x2).logrank_trafo2(x2)
x2 |
A vector Surv (Surv(time1, time2, event)) objects |
Logrank scores of LTRC objects
extree_data imports partykit::extree_data function
extree_data( formula, data, subset, na.action = stats::na.pass, weights, offset, cluster, strata, scores = NULL, yx = c("none", "matrix"), ytype = c("vector", "data.frame", "matrix"), nmax = c(yx = Inf, z = Inf), ... )extree_data( formula, data, subset, na.action = stats::na.pass, weights, offset, cluster, strata, scores = NULL, yx = c("none", "matrix"), ytype = c("vector", "data.frame", "matrix"), nmax = c(yx = Inf, z = Inf), ... )
formula |
Same as the one in extree_data, check extree_data for usage |
data |
Same as the one in extree_data, check extree_data for usage |
subset |
Same as the one in extree_data, check extree_data for usage |
na.action |
Same as the one in extree_data, check extree_data for usage |
weights |
Same as the one in extree_data, check extree_data for usage |
offset |
Same as the one in extree_data, check extree_data for usage |
cluster |
Same as the one in extree_data, check extree_data for usage |
strata |
Same as the one in extree_data, check extree_data for usage |
scores |
Same as the one in extree_data, check extree_data for usage |
yx |
Same as the one in extree_data, check extree_data for usage |
ytype |
Same as the one in extree_data, check extree_data for usage |
nmax |
Same as the one in extree_data, check extree_data for usage |
... |
Same as the one in extree_data, check extree_data for usage |
check extree_data for the return value
Recursive partition for interval-censored survival data in a conditional inference framework.
ICtree(Formula, data, Control = partykit::ctree_control())ICtree(Formula, data, Control = partykit::ctree_control())
Formula |
A formula object, with the response be a Surv object, with form Surv(time1, time2, type="interval2") |
data |
A data frame contains the variables named in Formula. |
Control |
A list of control parameters, see ctree_control |
ICtree returns a party object. This function extends
the conditional inference survival tree algorithm in ctree
to fit interval-censored survival data. This function itself not longer requires the interval package, but running the example below
requires the interval package (for bcos data), which in turn requires the Icens package, which is not available on CRAN. To install
the Icens package, enter the following commands
source("https://bioconductor.org/biocLite.R")
biocLite("Icens")
An object of class party.
Fu, W. and Simonoff, J.S. (2017). Survival trees for Interval Censored Survival data. Statistics in medicine 36 (30), 4831-4842
library(Icens) library(interval) library(LTRCtrees) data(bcos) ## Fit ICtree survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function Ctree <- ICtree(Surv(left,right,type="interval2")~treatment, data = bcos) ## Plot the fitted tree plot(Ctree)library(Icens) library(interval) library(LTRCtrees) data(bcos) ## Fit ICtree survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function Ctree <- ICtree(Surv(left,right,type="interval2")~treatment, data = bcos) ## Plot the fitted tree plot(Ctree)
LTRCART returns an rpart object. This function extends
the survival tree algorithm in rpart to fit left-truncated and
right censored (LTRC) data.
LTRCART( formula, data, weights = NULL, subset = NULL, no.SE = 0, control = rpart::rpart.control(cp = 0.001) )LTRCART( formula, data, weights = NULL, subset = NULL, no.SE = 0, control = rpart::rpart.control(cp = 0.001) )
formula |
A formula object specifies the regression function, with the response be a Surv object, with form Surv(time1, time2, event) |
data |
An optional data frame which contains the variables named in the formula. |
weights |
Optional case weights, same as in |
subset |
Optional expression saying that only a subset of the rows of the data should be
used in the fit, same as in |
no.SE |
Number of standard errors used in pruning, with default value 0. |
control |
A list of control values used to control the |
An object of class rpart. See rpart.object.
Fu, W. and Simonoff, J.S. (2017). Survival trees for left-truncated and right-censored data, with application to time-varying covariate data. Biostatistics 18 (2), 352-369.
## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Fit LTRCART survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function LTRCART.obj <- LTRCART(Surv(age, End, death) ~ sex + FLC + creatinine, Train) ## Putting Surv(End, death) in formula would result an error message ## since LTRCART is expecting Surv(time1, time2, event) ## Plot the fitted tree library(rpart.plot) rpart.plot(LTRCART.obj) ## Plot as partykit::party object library(partykit) plot(as.party(LTRCART.obj)) ## Plot as partykit::party object with survival curves on terminal nodes LTRCART.obj.party <- as.party(LTRCART.obj) LTRCART.obj.party$fitted[["(response)"]]<- Surv(Train$age, Train$End, Train$death) plot(LTRCART.obj.party) ## Predict relative risk on test set LTRCART.pred <- predict(LTRCART.obj, newdata = Test) #################################################################### ####### Survival tree with time-varying covariates ################## #################################################################### ## The pbcseq dataset of survival package library(survival) ## Create the start-stop-event triplet needed for coxph and LTRC trees first <- with(pbcseq, c(TRUE, diff(id) !=0)) #first id for each subject last <- c(first[-1], TRUE) #last id time1 <- with(pbcseq, ifelse(first, 0, day)) time2 <- with(pbcseq, ifelse(last, futime, c(day[-1], 0))) event <- with(pbcseq, ifelse(last, status, 0)) event <- 1*(event==2) pbcseq$time1 <- time1 pbcseq$time2 <- time2 pbcseq$event <- event ## Fit the Cox model and LTRCART tree with time-varying covariates fit.cox <- coxph(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) LTRCART.fit <- LTRCART(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) rpart.plot(LTRCART.fit) ### transform the wide format data into long format data using tmerge function ### from survival function ## Stanford Heart Transplant data jasa$subject <- 1:nrow(jasa) tdata <- with(jasa, data.frame(subject = subject, futime= pmax(.5, fu.date - accept.dt), txtime= ifelse(tx.date== fu.date, (tx.date -accept.dt) -.5, (tx.date - accept.dt)), fustat = fustat)) sdata <- tmerge(jasa, tdata, id=subject,death = event(futime, fustat), trt = tdc(txtime), options= list(idname="subject")) sdata$age <- sdata$age - 48 sdata$year <- as.numeric(sdata$accept.dt - as.Date("1967-10-01"))/365.25 Cox.fit <- coxph(Surv(tstart, tstop, death) ~ age+ surgery, data= sdata) LTRCART.fit <- LTRCART(Surv(tstart, tstop, death) ~ age + transplant, data = sdata) rpart.plot(LTRCART.fit)## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Fit LTRCART survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function LTRCART.obj <- LTRCART(Surv(age, End, death) ~ sex + FLC + creatinine, Train) ## Putting Surv(End, death) in formula would result an error message ## since LTRCART is expecting Surv(time1, time2, event) ## Plot the fitted tree library(rpart.plot) rpart.plot(LTRCART.obj) ## Plot as partykit::party object library(partykit) plot(as.party(LTRCART.obj)) ## Plot as partykit::party object with survival curves on terminal nodes LTRCART.obj.party <- as.party(LTRCART.obj) LTRCART.obj.party$fitted[["(response)"]]<- Surv(Train$age, Train$End, Train$death) plot(LTRCART.obj.party) ## Predict relative risk on test set LTRCART.pred <- predict(LTRCART.obj, newdata = Test) #################################################################### ####### Survival tree with time-varying covariates ################## #################################################################### ## The pbcseq dataset of survival package library(survival) ## Create the start-stop-event triplet needed for coxph and LTRC trees first <- with(pbcseq, c(TRUE, diff(id) !=0)) #first id for each subject last <- c(first[-1], TRUE) #last id time1 <- with(pbcseq, ifelse(first, 0, day)) time2 <- with(pbcseq, ifelse(last, futime, c(day[-1], 0))) event <- with(pbcseq, ifelse(last, status, 0)) event <- 1*(event==2) pbcseq$time1 <- time1 pbcseq$time2 <- time2 pbcseq$event <- event ## Fit the Cox model and LTRCART tree with time-varying covariates fit.cox <- coxph(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) LTRCART.fit <- LTRCART(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) rpart.plot(LTRCART.fit) ### transform the wide format data into long format data using tmerge function ### from survival function ## Stanford Heart Transplant data jasa$subject <- 1:nrow(jasa) tdata <- with(jasa, data.frame(subject = subject, futime= pmax(.5, fu.date - accept.dt), txtime= ifelse(tx.date== fu.date, (tx.date -accept.dt) -.5, (tx.date - accept.dt)), fustat = fustat)) sdata <- tmerge(jasa, tdata, id=subject,death = event(futime, fustat), trt = tdc(txtime), options= list(idname="subject")) sdata$age <- sdata$age - 48 sdata$year <- as.numeric(sdata$accept.dt - as.Date("1967-10-01"))/365.25 Cox.fit <- coxph(Surv(tstart, tstop, death) ~ age+ surgery, data= sdata) LTRCART.fit <- LTRCART(Surv(tstart, tstop, death) ~ age + transplant, data = sdata) rpart.plot(LTRCART.fit)
LTRCIT returns a party object. This function extends
the conditional inference survival tree algorithm in ctree
to fit left-truncated and right censored (LTRC) data.
LTRCIT(Formula, data, Control = partykit::ctree_control())LTRCIT(Formula, data, Control = partykit::ctree_control())
Formula |
A formula object, with the response be a Surv object, with form Surv(time1, time2, event) |
data |
A data frame contains the variables named in formula. |
Control |
A list of control parameters, see ctree_control |
An object of class party.
Fu, W. and Simonoff, J.S.(2017). Survival trees for left-truncated and right-censored data, with application to time-varying covariate data. Biostatistics 18 (2), 352-369.
## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Fit LTRCIT survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function LTRCIT.obj <- LTRCIT(Surv(age, End, death) ~ sex + FLC + creatinine, Train) plot(LTRCIT.obj) ## Putting Surv(End, death) in formula would result an error message ## since LTRCIT is expecting Surv(time1, time2, event) ## Note that LTRCIT.obj is an object of class party ## predict median survival time on test data LTRCIT.pred <- predict(LTRCIT.obj, newdata = Test, type = "response") ## predict Kaplan Meier survival curve on test data, ## return a list of survfit objects -- the predicted KM curves LTRCIT.pred <- predict(LTRCIT.obj, newdata = Test, type = "prob") #################################################################### ####### Survival tree with time-varying covariates ################## #################################################################### ## The pbcseq dataset of survival package library(survival) ## Create the start-stop-event triplet needed for coxph and LTRC trees first <- with(pbcseq, c(TRUE, diff(id) !=0)) #first id for each subject last <- c(first[-1], TRUE) #last id time1 <- with(pbcseq, ifelse(first, 0, day)) time2 <- with(pbcseq, ifelse(last, futime, c(day[-1], 0))) event <- with(pbcseq, ifelse(last, status, 0)) event <- 1*(event==2) pbcseq$time1 <- time1 pbcseq$time2 <- time2 pbcseq$event <- event pbcseq = pbcseq[1:1000,] ## fit on subset of the data to save fitting time ## Fit the Cox model and LTRCIT tree with time-varying covariates fit.cox <- coxph(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) LTRCIT.fit <- LTRCIT(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) plot(LTRCIT.fit) ## transform the wide format data into long format data using tmerge function ## from survival function ## Stanford Heart Transplant data jasa$subject <- 1:nrow(jasa) tdata <- with(jasa, data.frame(subject = subject, futime= pmax(.5, fu.date - accept.dt), txtime= ifelse(tx.date== fu.date, (tx.date -accept.dt) -.5, (tx.date - accept.dt)), fustat = fustat)) sdata <- tmerge(jasa, tdata, id=subject,death = event(futime, fustat), trt = tdc(txtime), options= list(idname="subject")) sdata$age <- sdata$age - 48 sdata$year <- as.numeric(sdata$accept.dt - as.Date("1967-10-01"))/365.25 Cox.fit <- coxph(Surv(tstart, tstop, death) ~ age+ surgery, data= sdata) LTRCIT.fit <- LTRCIT(Surv(tstart, tstop, death) ~ age + transplant, data = sdata) plot(LTRCIT.fit)## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Fit LTRCIT survival tree ## make sure to attach survival package (by library(survival) ) before using Surv function LTRCIT.obj <- LTRCIT(Surv(age, End, death) ~ sex + FLC + creatinine, Train) plot(LTRCIT.obj) ## Putting Surv(End, death) in formula would result an error message ## since LTRCIT is expecting Surv(time1, time2, event) ## Note that LTRCIT.obj is an object of class party ## predict median survival time on test data LTRCIT.pred <- predict(LTRCIT.obj, newdata = Test, type = "response") ## predict Kaplan Meier survival curve on test data, ## return a list of survfit objects -- the predicted KM curves LTRCIT.pred <- predict(LTRCIT.obj, newdata = Test, type = "prob") #################################################################### ####### Survival tree with time-varying covariates ################## #################################################################### ## The pbcseq dataset of survival package library(survival) ## Create the start-stop-event triplet needed for coxph and LTRC trees first <- with(pbcseq, c(TRUE, diff(id) !=0)) #first id for each subject last <- c(first[-1], TRUE) #last id time1 <- with(pbcseq, ifelse(first, 0, day)) time2 <- with(pbcseq, ifelse(last, futime, c(day[-1], 0))) event <- with(pbcseq, ifelse(last, status, 0)) event <- 1*(event==2) pbcseq$time1 <- time1 pbcseq$time2 <- time2 pbcseq$event <- event pbcseq = pbcseq[1:1000,] ## fit on subset of the data to save fitting time ## Fit the Cox model and LTRCIT tree with time-varying covariates fit.cox <- coxph(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) LTRCIT.fit <- LTRCIT(Surv(time1, time2, event) ~ age + sex + log(bili), pbcseq) plot(LTRCIT.fit) ## transform the wide format data into long format data using tmerge function ## from survival function ## Stanford Heart Transplant data jasa$subject <- 1:nrow(jasa) tdata <- with(jasa, data.frame(subject = subject, futime= pmax(.5, fu.date - accept.dt), txtime= ifelse(tx.date== fu.date, (tx.date -accept.dt) -.5, (tx.date - accept.dt)), fustat = fustat)) sdata <- tmerge(jasa, tdata, id=subject,death = event(futime, fustat), trt = tdc(txtime), options= list(idname="subject")) sdata$age <- sdata$age - 48 sdata$year <- as.numeric(sdata$accept.dt - as.Date("1967-10-01"))/365.25 Cox.fit <- coxph(Surv(tstart, tstop, death) ~ age+ surgery, data= sdata) LTRCIT.fit <- LTRCIT(Surv(tstart, tstop, death) ~ age + transplant, data = sdata) plot(LTRCIT.fit)
The output of LTRCART is an rpart object, and as a result the
usual predict function on such an object returns the predicted
relative risk on the test set. Pred.rpart returns the predicted
Kaplan-Meier curves and median survival times on the test set,
which in some circumstances might be desirable in practice.
Note that this function can be applied to any rpart survival tree
object, not just one produced by LTRCART
Pred.rpart(formula, train, test)Pred.rpart(formula, train, test)
formula |
A formula used to fit the survival tree. The response
is a Surv object. If it has the form Surv(time1, time2, event),
then |
train |
Training set |
test |
Test set |
A list of predicted KM curves and median survival times.
## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Predict median survival time and Kaplan Meier survival curve ## on test data using Pred.rpart LTRCART.pred <- Pred.rpart(Surv(age, End, death) ~ sex + FLC + creatinine, Train, Test) LTRCART.pred$KMcurves ## list of predicted KM curves LTRCART.pred$Medians ## vector of predicted median survival time## The Assay of serum free light chain data in survival package ## Adjust data & clean data library(survival) library(LTRCtrees) Data <- flchain Data <- Data[!is.na(Data$creatinine),] Data$End <- Data$age + Data$futime/365 DATA <- Data[Data$End > Data$age,] names(DATA)[6] <- "FLC" ## Setup training set and test set Train = DATA[1:500,] Test = DATA[1000:1020,] ## Predict median survival time and Kaplan Meier survival curve ## on test data using Pred.rpart LTRCART.pred <- Pred.rpart(Surv(age, End, death) ~ sex + FLC + creatinine, Train, Test) LTRCART.pred$KMcurves ## list of predicted KM curves LTRCART.pred$Medians ## vector of predicted median survival time