TKS: Threshold knot selection
In jelsema/RRSM: Methods for reduced rank spatial models

Description Usage Arguments Value See Also

Uses data to estimate the number and locations of knots for a reduced rank spatial process.

TKS(Y, X, coords, R, test_set = NULL, M = round(0.1 * nrow(X)),
  penalty = 0.5, BWs = NULL, num_bw = NULL, bw_int = c(0.03, 0.12),
  slices = NULL, num_slice = NULL, slice_int = c(0.98, 0.2),
  vpred = "EM", ...)

`Y`	the observed data. Should already be on log-scale.
`X`	the design matrix for the large-scale variation (fixed effects).
`coords`	the observed locations.
`R`	the residuals to use to select knots (computed if missing).
`test_set`	the indices of locations to use as the hold-out test set.
`M`	the number of locations to use for a hold-out test set (ignored if `test_set` is non-null).
`penalty`	the exponent in the penalty for the objective function of bandwidth selection. The MSE is weighted against the number of knots to this power.
`BWs`	the bandwidths to test.
`num_bw`	if `BWs` is null, this will set the number of bandwidths to test.
`bw_int`	if `BWs` is null, the maximum distance in the coordinates is multiplied by the min and max of `bw_int` to set the minimum and maximum bandwidths to test.
`slices`	the threshold cut-off values to define interval sets, defined as percentiles.
`num_slice`	if `slices` is null, this will set the number of percentile slices.
`slice_int`	if `slices` is null, the slices will be evenly spaces between the min and max of `slice_int`.
`vpred`	method to estimate parameters for the validation stage of TKS.
`...`	space for additional arguments.