Extract one candidate view from a grid fit — candidate

candidate_fit() turns one row of a grid-mode fit into a scalar-fit view. The returned object has the same class as object, with primary estimates and standard errors taken from object$tuning_grid, candidate-dependent extras such as W, ess, and neighbors rebuilt from cached ranks when the parent carried them, and candidate-independent state carried over. The rebuilt neighbors extra is the first k nearest observations, not the full DNN weight support.

Usage

candidate_fit(object, candidate, ...)

# Default S3 method
candidate_fit(object, candidate, ...)

# S3 method for class 'np_result'
candidate_fit(object, candidate, ...)

# S3 method for class 'cate_result'
candidate_fit(object, candidate, ...)

# S3 method for class 'plqr_result'
candidate_fit(object, candidate, ...)

candidate_fits(object, candidates = "all", ...)

# Default S3 method
candidate_fits(object, candidates = "all", ...)

# S3 method for class 'np_result'
candidate_fits(object, candidates = "all", ...)

# S3 method for class 'cate_result'
candidate_fits(object, candidates = "all", ...)

# S3 method for class 'plqr_result'
candidate_fits(object, candidates = "all", ...)

Arguments

object: A grid fit returned by np_dnn(), cate_dnn(), or plqr_dnn() with tuning_mode = "grid".
candidate: A single candidate_id string or integer row index into object$tuning_grid$candidates.
...: Reserved for future methods.
candidates: "all" (default), a character vector of candidate ids, or an integer vector of candidate row indices.

Value

A scalar-style np_result, cate_result, or plqr_result view.

A named list of scalar-style candidate views.

Experimental methods

Methods for plqr_result objects are experimental, as is all PLQR functionality in this release; see plqr_dnn().

Examples

set.seed(1)
x_mat <- matrix(rnorm(40 * 2), 40, 2)
y_vec <- x_mat[, 1] + rnorm(40, sd = 0.2)
fit <- np_dnn(
  X = x_mat, x0 = c(0, 0), Y = y_vec,
  tuning_mode = "grid", se = FALSE
)
candidate_fit(fit, fit$tuning_grid$candidates$candidate_id[[1L]])
#> <np_result>
#>   data  : n=40, p=2, m=1
#>   DNN   : method=exact, k=1, s=4, distance=l2
#>   est   : mu[1:min(5,m)] = 0.0386  (SEs not requested)