Normalize Outcome/Response to [0,1] Interval

Description

This function takes a continuous (double) column of data and converts it to have 0 as the lower bound and 1 as the upper bound.

Usage

normalize(outcome, true_bounds = NULL)

Arguments

outcome	Any non-character vector. Factors will be converted to numeric via coercion.
true_bounds	Specify this parameter with the lower and upper bound if the observed min/max of the outcome should not be used. Useful when an upper or lower bound exists but the observed data is less than/more than that bound. The normalization function will respect these bounds.

Details

Beta regression can only be done with a response that is continuous with a lower bound of 0 and an upper bound of 1. However, it is straightforward to transform any lower and upper-bounded continuous variable to the [0,1] interval. This function does the transformation and saves the original bounds as attributes so that the bounds can be reverse-transformed.

Value

A numeric vector with an upper bound of 1 and a lower bound of 0. The original bounds are saved in the attributes "lower_bound" and "upper_bound".

Examples

library(ordbetareg)

# set up arbitrary upper and lower-bounded vector
outcome <- runif(1000, min=-33, max=445)

# normalize to \[0,1\]

trans_outcome <- normalize(outcome=outcome)
summary(trans_outcome)

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
 0.0000  0.2655  0.4860  0.4951  0.7367  1.0000 

# only works with numeric vectors and factors

  try(normalize(outcome=c('a','b')))

Error in normalize(outcome = c("a", "b")) : 
  Please do not pass a character vector as a response/outcome.
That really doesn't make any sense.