PCA Component as a LaTeX Linear Combination

Writes one principal component as a LaTeX sum of signed loadings times feature names, for a report (for example PC_1 = +0.412 x - 0.231 y). Loadings are the elements of `rotation` (signed; the sign of a component is arbitrary). Underscores and other LaTeX specials in feature names are escaped.

Usage

pca_equation_latex(
  pca_mod,
  pc = 1,
  feature_names = NULL,
  max_terms = NULL,
  prefix = "PC",
  digits = 3
)

Arguments

pca_mod

A [stats::prcomp()] object or a [run_pca()] result.

pc

Component number or name (`"PC2"`); exactly one.

feature_names

Optional display names for the features, either one per rotation row (in order) or a named vector (names = original features) that may cover only some of them. `NULL` uses the row names of the rotation.

max_terms

Keep only this many terms, those with the largest absolute loading (shown in that order, followed by `...`); `NULL` keeps all in the original order.

prefix

Component label prefix.

digits

Decimals of the loadings.

Value

A single string of LaTeX math, without delimiters (wrap it in `$$ ... $$`).

See also

[pca_component_cutoff()], [pca_top_features()], [pca_eqns()]

Examples

fit <- prcomp(mtcars[, 1:4], scale. = TRUE)
pca_equation_latex(fit, 1, digits = 2)
#> [1] "PC_{1} = -0.50 \\text{mpg} +0.51 \\text{cyl} +0.51 \\text{disp} +0.48 \\text{hp}"
pca_equation_latex(fit, 2, max_terms = 2)
#> [1] "PC_{2} = +0.848 \\text{hp} +0.415 \\text{mpg} \\dots"