Get the target of a fit

get_target() returns the target that the estimates of a fit come from. For a stack_direct fit this is the BRR-Fay target from pv_brr_target(), to which the stacked fit was calibrated; for a per_pv fit it is the Rubin's-rules combination of the draws of the per-plausible-value fits. A stack_psis fit has no target object.

Usage

get_target(x, ...)

# Default S3 method
get_target(x, ...)

# S3 method for class 'pvstackr_fit'
get_target(x, ...)

# S3 method for class 'pvstackr_brr_target'
get_target(x, ...)

# S3 method for class 'pvstackr_legacy_psis_inspection'
get_target(x, ...)

Arguments

x

A fit (class pvstackr_fit), or a target from pv_brr_target().

...

Ignored.

Value

For a stack_direct fit, its pvstackr_brr_target object, also when the fit is blocked. pv_brr_target() lists its elements, such as beta, T_MI, df and fmi (the fraction of missing information, which the estimate table does not have). For a per_pv fit, a pvstackr_reference_pool object with the same kind of elements. For a stack_psis fit, NULL: its estimate table has target_source = "stack_psis_rubin_pooling", but that is only a label for the combined result, not a target object. Given a target, get_target() checks it and returns it unchanged.

get_target() stops with an error if the fit or target was changed after it was created. For the inspection object that pv_migrate_legacy_psis_fit() makes from a stack_psis fit of an earlier pvstackr version, it returns NULL.

See also

pv_brr_target() for the elements of a target.

Other pvstackr-accessors: get_diagnostics(), get_draws(), get_estimates()

Examples

path <- system.file(
  "extdata", "examples", "pisa_tiny_stack_direct.rds", package = "pvstackr"
)
if (nzchar(path)) {
  fit <- readRDS(path)$fit
  get_target(fit)
}
#> pvstackr BRR-Fay target
#>   fixed effects: 3
#>   plausible values: 2
#>   replicate weights: 4
#>   fay_k: 0.5
#>   df method: classic
#>   interval role: descriptive_classic_rubin
#>   source: external_brr_fay_rubin