approx_dt: Approximate missing values in a data.table.
In pik-piam/rmndt: Tools for data.table objects in the REMIND context

approx_dt

R Documentation

Approximate missing values in a data.table.

Description

Similar to, but not quite like, 'stats::approx'. Does only support constant extrapolation and linear interpolation. The resulting 'data.table' only contains the range provided by 'xdata' along 'xcol'. Without extrapolation, 'xcol' in the resulting 'data.table' may not cover the range given by 'xdata'.

Usage

approx_dt(
  dt,
  xdata,
  xcol,
  ycol,
  idxcols = NULL,
  keepna = FALSE,
  extrapolate = FALSE
)

Arguments

`dt`	a data.table.
`xdata`	the range to interpolate to. This is the range the result will have along the dimension 'xcol'.
`xcol`	name of the column for interpolation.
`ycol`	name of the column that contains the value to be interpolated.
`idxcols`	columns that identify a row (besides xcol), i.e., the remaining index dimensions.
`keepna`	keep NA values for rows that can not be interpolated (since they are outside of [min(xcol), max(xcol)]), default is FALSE.
`extrapolate`	use the closest values to fill 'ycol' outside of the interpolation domain, default is FALSE. This will also work if there is only one value along 'ycol', i.e., no interpolation is taking place.

Value

a data.table with the range given by 'xdata' along 'xcol'. Columns not given in 'idxcols' will be kept but NAs will appear on extrapolated and interpolated rows.

Examples

dt <- as.data.table(ChickWeight)
## delete all values but 1
dt[Chick == 1 & Time > 0, weight := NA]
## delete all values but 2
dt[Chick == 2 & Time > 2, weight := NA]

## extrapolation from 1 value
approx_dt(dt, 0:21, "Time", "weight", idxcols=c("Chick", "Diet"), extrapolate = TRUE)[Chick == 1]
## extrapolation and interpolation
approx_dt(dt, 0:21, "Time", "weight", idxcols=c("Chick", "Diet"), extrapolate = TRUE)[Chick == 2]
## column not in idxcols
approx_dt(dt, 0:21, "Time", "weight", idxcols="Chick", extrapolate = TRUE)[Chick == 2]

dt <- as.data.table(ChickWeight)
## interpolation only
approx_dt(dt, 0:21, "Time", "weight", idxcols=c("Chick", "Diet"))[Chick == 2]

pik-piam/rmndt documentation built on April 21, 2024, 4:31 a.m.