Create a non-linear (sigmoidal) scoring rule

Creates a scoring rule object for indicators to be scored with the sigmoidal curve \(S = 1 / (1 + (x/x_0)^b)\) rather than with linear min-max normalization. See score_sigmoid for the full description of the curve, the meaning of b, and the caveat that the literature does not agree on whether the linear or non-linear form is preferable.

Usage

sigmoid_scoring(direction = c("higher", "lower"), x0 = NULL, b = 2.5)

Arguments

direction

Either "higher" (more is better, the default) or "lower" (less is better).

x0

Optional reference value at which the score equals 0.5. If NULL, the mean of the indicator is used when the rule is applied. Supplying an external reference is what makes scores comparable across studies.

b

Shape parameter controlling steepness. Must be positive; the direction argument supplies the sign. Defaults to 2.5, a convention inherited from the literature rather than a general constant.

Value

A scoring_rule object of class c("scoring_rule", "sigmoid_scoring")

See also

score_sigmoid for the underlying function; higher_better, lower_better for the linear equivalents; standard_scoring_rules to generate sigmoidal rules for a whole property set at once

Examples

# Organic matter, scored non-linearly against the sample mean
om_rule <- sigmoid_scoring()

# Bulk density, anchored on an external reference value
bd_rule <- sigmoid_scoring(direction = "lower", x0 = 1.4)

# A sharper curve
om_rule <- sigmoid_scoring(b = 6)