Skip to contents

Control Parameters for the Scorecard Pipeline

Source: R/obwoe_scorecard.R
control.obwoe_scorecard.Rd

Settings for the stages of obwoe_scorecard that have no home in control.obwoe or obwoe_select.

Usage

control.obwoe_scorecard(
  corr_cutoff = 0.7,
  corr_method = "pearson",
  pdo = 20,
  score_ref = 600,
  odds_ref = 50,
  direction = "higher_is_safer",
  n_groups = 10L,
  na_woe = 0,
  drop_negative = TRUE,
  max_abs_coef = 15,
  min_events_per_variable = 20,
  engine_fallback = FALSE,
  overwrite = TRUE,
  digits = 6L
)

Arguments

corr_cutoff

Numeric in \((0, 1]\). Absolute correlation in the WoE space at or above which two variables are treated as redundant. Default 0.70. Set to 1 to disable pruning.

corr_method

Character string passed to obcorr. Default "pearson".

pdo, score_ref, odds_ref, direction

Passed to obwoe_scale.

n_groups

Integer. Number of score bands in the gains and stability tables. Default 10.

na_woe

Numeric. WoE given to a value that falls in no fitted bin — an unseen category, or a missing value where the binner built no missing bin. Default 0, which is the population-average contribution.

drop_negative

Logical. Refit without any variable whose coefficient on WoE comes out negative? Default TRUE. A negative coefficient means the variable is fighting the rest of the model, and a scorecard with one cannot be signed off.

max_abs_coef

Numeric. Coefficients above this in absolute value are flagged as separation. Default 15.

min_events_per_variable

Numeric. Events per variable in the model below which the fit is flagged as thin. Default 20, the usual rule of thumb.

engine_fallback

Logical. If the requested engine's package is not installed, fall back to "glm" with a warning instead of failing? Default FALSE.

overwrite

Logical. Overwrite file if it exists? Default TRUE.

digits

Integer. Rounding used in the workbook. Default 6.

Value

An object of class "obwoe_scorecard_control".

See also

obwoe_scorecard

Examples

control.obwoe_scorecard(pdo = 25, score_ref = 700, corr_cutoff = 0.8)
#> $corr_cutoff
#> [1] 0.8
#> 
#> $corr_method
#> [1] "pearson"
#> 
#> $pdo
#> [1] 25
#> 
#> $score_ref
#> [1] 700
#> 
#> $odds_ref
#> [1] 50
#> 
#> $direction
#> [1] "higher_is_safer"
#> 
#> $n_groups
#> [1] 10
#> 
#> $na_woe
#> [1] 0
#> 
#> $drop_negative
#> [1] TRUE
#> 
#> $max_abs_coef
#> [1] 15
#> 
#> $min_events_per_variable
#> [1] 20
#> 
#> $engine_fallback
#> [1] FALSE
#> 
#> $overwrite
#> [1] TRUE
#> 
#> $digits
#> [1] 6
#> 
#> attr(,"class")
#> [1] "obwoe_scorecard_control"