Settings for frame() — framedf

Returns the list of tunable thresholds used by frame(). Every field can be overridden through the ... argument of frame(). The defaults are chosen to be conservative on small data and to scale to millions of rows through progressive subsampling.

Usage

framedf_settings(adjustment = NULL, ...)

Arguments

adjustment: Optional adjustment columns (passed through unchanged).
...: Named overrides.

Value

A named list of settings.

Fields

min_obs: Minimum complete cases required to screen a pair.
strong_threshold, moderate_threshold, weak_threshold: Absolute correlation cut-offs used to classify pair strength.
near_constant_ratio: Maximum allowed share of the dominant value before a column is flagged as near-constant.
id_unique_ratio: If a character column has more than this share of unique values, it is treated as an identifier.
skew_threshold: Absolute skewness above which a column is flagged as right- or left-skewed.
subsample_threshold: Row count above which numeric pairs are screened by progressive subsampling.
subsample_probe, subsample_confirm: Probe and confirmation sample sizes for the two-stage subsampling.
compositional_cv: Maximum coefficient of variation of x + y for the pair to be flagged as a constrained complement.
observer_min_levels: Minimum number of categorical levels before a strong group effect counts as an observer-style concern.
min_level_n: Levels with fewer than this many rows are flagged as rare.
seed: Seed used by the subsampling layer.
adjustment: Optional column names to partial out before numeric pair screening.