Skip to contents

Select smoothing parameters for circular-linear conditional density

Source: R/directional-smoothing.R
select_cyl_smoothing.Rd

Ranks candidate smoothing parameters by K-fold conditional log-likelihood.

Usage

select_cyl_smoothing(
  theta,
  x,
  kappa_values = c(6, 8, 12, 16, 24, 36),
  h_values = NULL,
  n_folds = 5,
  seed = NULL
)

Arguments

theta

Conditioning angles in radians.

x

Linear response.

kappa_values

Candidate circular concentration values.

h_values

Candidate linear bandwidth values. If NULL, a small grid is built around the rule-of-thumb bandwidth used by estimate_cyl_density().

n_folds

Number of cross-validation folds.

seed

Optional random seed for fold assignment.

Value

A data frame ranked by decreasing mean held-out log score. The returned object has class directional_smoothing_selection.

See also

Other directional smoothing helpers: select_toroidal_smoothing()

Examples

dat <- simulate_cylindrical(n = 80, seed = 1)
select_cyl_smoothing(
  dat$theta, dat$x,
  kappa_values = c(8, 12),
  h_values = c(0.3, 0.5),
  n_folds = 2,
  seed = 1
)
#>   kappa   h mean_log_score se_log_score rank
#> 1    12 0.5      -1.160393   0.03815670    1
#> 2     8 0.5      -1.184392   0.03798779    2
#> 3     8 0.3      -1.306153   0.10941871    3
#> 4    12 0.3      -1.314517   0.10744874    4