Skip to contents

Clean Counting-Process Intervals

Source: R/surv_helpers.R
prune_start_stop.Rd

Companion to [validate_start_stop()]: drops intervals with `stop <= start`, drops rows recorded after a subject's first event, and closes tiny gaps or overlaps (smaller than `tol`) between a subject's consecutive intervals by moving the later `start` to the earlier `stop`. Rows with a missing id, time or status are dropped too. Gaps and overlaps larger than `tol` are real and left alone; check them with [validate_start_stop()].

Usage

prune_start_stop(
  data,
  id = "id",
  start = "tstart",
  stop = "tstop",
  status = "status",
  tol = 1e-06
)

Arguments

data

A data frame of counting-process intervals.

id, start, stop, status

Column names (strings). Status is 0/1, logical or a 0/1 factor.

tol

Largest gap or overlap treated as rounding error (default `1e-6`).

Value

`data` with the offending rows removed, sorted by `id`, `start`, `stop`. The counts are in `attr(x, "pruned")`: `missing`, `nonpositive`, `after_event` and `gaps_closed`.

See also

[validate_start_stop()]

Examples

df <- data.frame(id = c(1, 1, 1, 2), tstart = c(0, 5.0000001, 10, 0),
                 tstop = c(5, 10, 12, 0), status = c(0, 1, 0, 0))
prune_start_stop(df)
#>   id tstart tstop status
#> 1  1      0     5      0
#> 2  1      5    10      1