Skip to contents

Validate contrast weights against a design matrix or event model

Source: R/validate.R
validate_contrasts.Rd

Provides basic diagnostics for t- and F-contrasts once the design matrix is available. You can either pass an event_model (to validate all attached contrasts) or a design matrix plus custom weights.

Usage

validate_contrasts(x, weights = NULL, tol = 1e-08)

Arguments

x

An event_model or a numeric matrix/data.frame design matrix.

weights

Optional contrast weights. May be a numeric vector (t-contrast), a numeric matrix (F-contrast with columns as contrast vectors), or a named list mapping names to vectors/matrices. If NULL and x is an event_model, all attached t- and F-contrasts are validated.

tol

Numeric tolerance for zero checks. Default 1e-8.

Value

A data.frame with one row per validated contrast column and the following columns: name, type ("t" or "F"), estimable, sum_to_zero, orthogonal_to_intercept, full_rank (F only), and nonzero_weights.

Details

Checks include:

  • Estimability: whether each contrast column lies in the row space of X.

  • Sum-to-zero: whether the weights sum to ~0 (t-contrasts only).

  • Intercept orthogonality: whether weights on intercept-like columns are ~0.

  • Full-rank (F only): whether an F-contrast matrix has full column rank.

Examples

if (FALSE) { # \dontrun{
# Validate all attached contrasts on a model
res <- validate_contrasts(emodel)

# Validate a custom vector against a model
v <- rep(0, ncol(design_matrix(emodel))); v[1] <- 1; v[2] <- -1
res2 <- validate_contrasts(emodel, weights = v)

# Validate a custom matrix against a design matrix
X <- as.matrix(design_matrix(emodel))
C <- cbind(c(1,-1,rep(0, ncol(X)-2)), c(0,1,-1,rep(0, ncol(X)-3)))
res3 <- validate_contrasts(X, weights = C)
} # }