Temporal sampling

Divides a time window into equal intervals and samples within each one. per_interval is named for what it is: unlike n elsewhere in the package, it is a per-group figure, not a total.

Usage

design_temporal(
  time,
  from,
  to,
  interval,
  per_interval,
  unit = c("hours", "seconds", "minutes", "days", "weeks", "months", "years"),
  tz = NULL,
  na_rm = FALSE
)

Arguments

time

Column of timestamps. POSIXct and Date columns are used as they are; character columns are parsed with lubridate::ymd_hms().

from, to

Window bounds, as timestamps or strings. The window is half-open: [from, to). Trailing partial intervals are dropped.

interval

Interval width, in units of unit.

per_interval

Rows to draw from each interval. Intervals holding fewer rows contribute all of them.

unit

One of "seconds", "minutes", "hours", "days", "weeks", "months" or "years". "months" and "years" step by calendar units, not by fixed 30.44-day durations.

tz

Time zone the timestamps are expressed in. Applied when the column and both bounds are parsed, so it genuinely shifts the window rather than relabelling the output.

na_rm

Drop rows whose timestamp is missing or unparseable instead of raising an error.

Value

A design object, for use with draw().

Details

Buckets are assigned arithmetically in a single pass rather than by re-filtering the data per interval, so cost scales with the data rather than with the number of intervals.

See also

draw()

Other designs: design_bootstrap(), design_certainty(), design_cluster(), design_multistage(), design_reservoir(), design_simple(), design_spatial(), design_spread(), design_stratified(), design_systematic(), design_weighted()

Examples

df <- data.frame(
  id = 1:48,
  ts = seq(as.POSIXct("2020-01-01", tz = "UTC"), by = "hour", length.out = 48)
)
d <- design_temporal("ts", from = "2020-01-01", to = "2020-01-02",
                     interval = 6, per_interval = 2, unit = "hours")
draw(df, d, seed = 1)$ts
#> [1] "2020-01-01 00:00:00 UTC" "2020-01-01 03:00:00 UTC"
#> [3] "2020-01-01 06:00:00 UTC" "2020-01-01 07:00:00 UTC"
#> [5] "2020-01-01 14:00:00 UTC" "2020-01-01 16:00:00 UTC"
#> [7] "2020-01-01 19:00:00 UTC" "2020-01-01 23:00:00 UTC"