Check Affine Normalization (Zero-Centered Min-Max Scaling)

Verifies that a numeric vector has been affine normalized such that it is zero-centered and its minimum value is at least -1. Specifically, the function checks that:

Usage

check_affine_normalization(x, verbose = FALSE)

Arguments

x

A numeric vector that has been affine normalized. This vector may contain NA values.

verbose

Logical; if TRUE, the function prints details of the check to the console.

Value

Invisibly returns NULL if the check passes. Otherwise, the function stops with an error message detailing the failed condition(s).

Details

• The mean of the vector (ignoring NAs) is approximately zero.

• The minimum value (ignoring NAs) is greater than or equal to -1.

• The maximum value (ignoring NAs) is above zero.

If any of these conditions fail, the function stops execution and returns an informative error message.

The check uses a tolerance of 1e-2: the mean must satisfy abs(mean) < 1e-2 and the minimum must be at least -1 - 1e-2. The conditions verified correspond to a transformation defined as:

$$x_{\text{scaled}} = \frac{x - \mu}{\mu - \min(x)}$$

where \(\mu\) is the mean of x. Note that while the transformation forces a zero-centered result, the minimum value need not be exactly -1 but should be no less than -1.

Examples

# Example vector after affine normalization
x <- c(-1, 0, 0.5, 1.2, NA)

# Check without printing details:
if (FALSE) { # \dontrun{
  check_affine_normalization(x)
} # }

# Check with verbose output:
if (FALSE) { # \dontrun{
  check_affine_normalization(x, verbose = TRUE)
} # }