Scaling range transformation function (re-scaling)

This function supports multiple scaling methods to transform risk values from an existing range to another. By default the output range is between 0 and 100.

Usage

rescale_risk_scores(
  dataset,
  cols = NULL,
  from = NULL,
  to = c(0, 100),
  method = c("linear", "quadratic", "exponential", "sigmoid"),
  inverse = FALSE,
  names_prefix = NULL,
  names_to = NULL,
  keep_cols = FALSE
)

Arguments

dataset

dataset to add scaled risk column to.

cols

Column containing numeric vector of risk values in the range [0, 12] OR a named vector, names will be used as the names_to arg, giving new names to rescaled columns.

from

existing range of possible values for risk_col that will be converted to

to

new range of possible values for risk_col, by default it is 0 to 100.

method

The scaling method to apply. Options are "linear", "quadratic", "exponential", "sigmoid", or "complementary".

inverse

boolean to inverse the risk values, i.e. low becomes high, and high becomes low. Similar to the "complementary" value of method, but can be added to quadratic.

names_prefix

string, prefix cols names to have new names for scaled columns.

names_to

string vector, provide new name for rescaled columns.

keep_cols

default TRUE, whether to keep cols columns after rescaling is done or to remove.

Value

A dataset with new column containing numeric vector of scaled values in the range [0, 100].

Details

Below are the equations used for each method:

Linear Scaling $$f(x) = \frac{x - 0}{12 - 0} \cdot (100 - 0) + 0$$

Quadratic Scaling $$f(x) = \left(\frac{x - 0}{12 - 0}\right)^2 \cdot (100 - 0) + 0$$

Exponential Scaling $$f(x) = \frac{\exp(x / 12) - 1}{\exp(1) - 1} \cdot (100 - 0) + 0$$

Sigmoid Scaling $$f(x) = \frac{100}{1 + \exp(-10 \cdot (\frac{x}{12} - 0.5))}$$

Examples

# rescale_risk -----
library(riskintrodata)
library(riskintroanalysis)

dat <- data.frame(RISK = seq(0, 12, by = 0.01))

dat_scaled <- rescale_risk_scores(
  dataset = dat,
  cols = "RISK",
  method = "quadratic",
  from = c(0, 12),
  to = c(0, 100),
  names_prefix = "quadratic_",
  keep_cols = TRUE
)

ggplot(dat_scaled, aes(x = RISK, y  = quadratic_RISK)) +
  geom_point()