Skip to contents

Transform Exposure Variables for Normality

Source: R/transform_exposure.R
transform_exposure.Rd

Applies a transformation to selected numeric exposure variables in the colData of a MultiAssayExperiment to improve their normality (e.g., log, Box-Cox, sqrt). Transformation results and normality statistics are stored in metadata for tracking.

Usage

transform_exposure(expomicset, exposure_cols = NULL, transform_method = "best")

Arguments

expomicset

A MultiAssayExperiment object containing exposure variables in colData.

exposure_cols

Optional character vector of exposure variable names to transform. If NULL, uses exposures found in metadata(expomicset)$quality_control$normality$norm_df$exposure.

transform_method

Character. Transformation method to apply. Options:

  • "none": no transformation

  • "log2": log base 2 transformation

  • "x_1_3": cube-root transformation

  • "sqrt": square-root transformation

  • "boxcox_best": data-driven Box-Cox approximation with heuristic labeling

Value

A MultiAssayExperiment object with transformed exposures in colData, and transformation details stored in:

  • metadata(expomicset)$quality_control$transformation$norm_df: Shapiro-Wilk test results

  • metadata(expomicset)$quality_control$transformation$norm_summary: Summary of normality

  • metadata(expomicset)$codebook: Updated with transformation info per variable

  • metadata(expomicset)$summary$steps: Updated with step record

Details

For transform_method = "boxcox_best", the function automatically shifts values to be strictly positive and chooses from a discrete set of transformations (e.g., 1/x, log(x), x^2) based on estimated Box-Cox lambda. Each variable may receive a different transformation.

See also

boxcox, shapiro.test

Examples

if (FALSE) { # \dontrun{
expomicset <- transform_exposure(expomicset, transform_method = "boxcox_best")
} # }