Episode splitting with qualitative covariates

gistfile1.r

## Note: No missing values in tvc.start_/tvc.stop_ allowed!


## 0. Step: Make up some data
##          event:      event status
##          start:      starting time
##          stop:       ending time
##          tvc.start_: starting time of qualitative/categorial TVC
##          tvc.stop_:  ending time of qualitative/categorial TVC
df <- data.frame(id=c(1, 2),
                 event=c(1, 0),
                 start=c(0, 0),
                 stop=c(12, 8),
                 tvc.start1=c(2, 4),
                 tvc.start2=c(3, 7),
                 tvc.stop1=c(12, 7),
                 tvc.stop2=c(7, 8)
                 )
df


## 1. Step: Reshape data frame
##          ALL potential endpoints need to be collapsed into one vector and,
##          by default, R uses the first name in varying=list(...) for the
##          name of this newly created vector (here: "stop")
data.long <- reshape(data=df, direction="long", timevar="index",
                     varying=list(c("stop", "tvc.start1", "tvc.stop1",
                     "tvc.start2", "tvc.stop2")))
data.long
## Remove "index" variable (see reshape(..., timevar="index"))
data.long <- data.long[, names(data.long)!="index"]



## 2. Step: Sort data by id and event time (here: "stop")
data.long <- data.long[order(data.long$id, data.long$stop), ]
data.long



## 3. Step: Correct starting times (i.e. generate left side of intervals)
##          Don't worry, this is tricky step and I do not expect you to
##          understand this step...
tmp.list <- split(data.long[,c("id", "stop")], data.long$id)
tmp.list.start <- lapply(tmp.list, function(x){c(0, x$stop[1:length(x$stop)-1])})
data.long$start <- unlist(tmp.list.start)
data.long



## 4. Step: If end of a TVC episode >= end of observation time,
##          remove double record
data.long <- data.long[!data.long$start==data.long$stop, ]
data.long



## 5. Step: Merge long and wide data
##          Note: variable "stop" exists in both data sets
data.long <- merge(data.long, df[, c("id", "stop", "tvc.start1", "tvc.start2",
                                     "tvc.stop1", "tvc.stop2")], by="id",
                   all.x=TRUE)
names(data.long)[names(data.long)=="stop.x"] <- "stop"
data.long



## 6. Step: Correct old "event" status variable
data.long$event2 <- data.long$event
data.long$event <- 0
data.long$event[(data.long$stop == data.long$stop.y)
                & (data.long$event2 == 1)] <- 1
data.long



## 7. Step: Generate TVC dummies
data.long$tvc.dummy1 <- 0
data.long$tvc.dummy1[data.long$start >= data.long$tvc.start1
                     & data.long$stop <= data.long$tvc.stop1] <- 1
data.long

data.long$tvc.dummy2 <- 0
data.long$tvc.dummy2[data.long$start >= data.long$tvc.start2
                     & data.long$stop <= data.long$tvc.stop2] <- 1
data.long