Impute missing data points — spc_na

Replace NAs in the spectra matrix by interpolation. With less than 4 points available linear interpolation of the 2 neighbor points is used. For larger numbers of neighbor points, smoothing interpolation is performed by stats::smooth.spline().

spc_na_approx(
  spc,
  neighbours = 1,
  w = rep(1, 2 * neighbours),
  df = 1 + .Machine$double.eps,
  spar = NULL,
  debuglevel = hy.getOption("debuglevel")
)

Arguments

spc	hyperSpec object with spectra matrix containing `NA`s
neighbours	how many neighbor data points should be used to fit the line
w, df, spar	see `stats::smooth.spline()`
debuglevel	see `options()`

Value

hyperSpec object

Author

Claudia Beleites

Examples

fluNA <- hyperSpec:::fluNA
spc_na_approx(fluNA[, , min ~ 410], debuglevel = 1)
#> Warning: Spectra containing only NAs found. They will not be approximated.
#> hyperSpec object
#>    6 spectra
#>    3 data columns
#>    11 data points / spectrum
spc_na_approx(fluNA[1, , min ~ 410], debuglevel = 2)
#> hyperSpec object
#>    1 spectra
#>    3 data columns
#>    11 data points / spectrum
spc_na_approx(fluNA[4, , min ~ 410], neighbours = 3, df = 4, debuglevel = 2)
#> hyperSpec object
#>    1 spectra
#>    3 data columns
#>    11 data points / spectrum