R: Inverse logit is robust to cases that equal 0 or 1

invLogitAdj {statisfactory}

R Documentation

Inverse logit is robust to cases that equal 0 or 1

Description

This function is the inverse of logitAdj. That function calculates the logit of values but is robust to cases where the operand is 0 or 1. The adjusted inverse logit is equal to (base^x + epsilon * base^x - epsilon) / (base^x + 1).

Usage

invLogitAdj(x, epsilon = 0.01, base = 10, auto = FALSE)

Arguments

`x`	Numeric vector.
`epsilon`	Value or character. If a numeric value (typically ~0.01 or smaller), then this is added/subtracted from `x` to ensure log of 0 or 1 is not taken. If equal to `'auto'` then the value of `epsilon` is taken from the attributes of `x`. If `x` has no such attribute, a warning is given and a value of 0.01 is used.
`base`	Base of logarithm. Use `base=exp(1)` for base `e`.
`auto`	If `TRUE` then if the attributes of `x` have slots named `epsilon` and `base` then use these instead of the user-supplied values of `epsilon` and `base`. If they do not appear as attributes of `x` but `auto` is `TRUE` then the function prints warnings and uses 0.01 and 10, respectively. If `FALSE` (default) then use the user-supplied values of `epsilon` and `base`.

Value

Numeric.

Examples


x <- seq(0, 1, by=0.1)
y <- logitAdj(x)
xx <- invLogitAdj(y, auto = TRUE)