buildQuasiObsForTraining: Expand discrete time survival data into quasi observations...
In WinVector/QSurvival: Quick survival models by data transformation

Description Usage Arguments Examples

This function assumes time starts at 1 and counts unit increments.

1
2
3

buildQuasiObsForTraining(d, numberOfObservations, eventIndex, idColumnName,
  indexColumnName, eventColumnName, parallelCluster = NULL, targetSize = NA,
  forceEvent = FALSE, weightsColumnName = NULL)

`d`	data.frame to expand
`numberOfObservations`	vector (entries >=1) for each row of d up to what step number did we observe
`eventIndex`	vector for each row of d at what time did event occur. Event i is considered censored if eventIndex[[i]] is NA or out of range of 1:numberOfObservations.
`idColumnName`	character scalar, column to write original row ids in
`indexColumnName`	character scalar, column to write quasi event indices into
`eventColumnName`	character scalar, column to write quasi events indicator into. If NA, write no such column.
`parallelCluster,`	if not NULL parallel cluster to perform the work.
`targetSize`	numeric scalar, if not NA target size for a uniform sample of the quasi observations. This should be a very large number.
`forceEvent`	logical scalar, if true force conversion events into sample
`weightsColumnName`	if not NULL column to write sampling weights in

1
2
3

d <- data.frame(lifetime=c(2,1,2),censored=c(FALSE,FALSE,TRUE))
buildQuasiObsForTraining(d,d$lifetime,ifelse(d$censored,NA,d$lifetime),
   'origRow','sampleAge','deathEvent')