ancova_linear_contrast: Compute a single ANCOVA linear contrast

In rbmi: Reference Based Multiple Imputation

Compute a single ANCOVA linear contrast

Description

Evaluates a linear contrast over the group levels as a linear combination of the model coefficients. The group-level weights are applied to complete model design rows evaluated at the covariate reference point, so interactions are included and the result is independent of the active contrasts coding.

Usage

ancova_linear_contrast(weights, beta, vcov_mod, df_res, contrast_design)

Arguments

weights	Numeric weight vector over the group levels (factor order), summing to zero.
beta	Numeric vector of model coefficients.
vcov_mod	Variance-covariance matrix of the model coefficients.
df_res	Residual degrees of freedom.
contrast_design	Numeric model matrix with one row per group level, evaluated at the covariate reference point, and one column per model coefficient.

Details

The contrast is evaluated at the reference values defined by ancova_contrast_design(). Finite columns involving only non-group covariates are identical across group rows and cancel for a sum-to-zero contrast. Group-by-covariate interaction columns are included when they are non-zero at that reference point, although ancova() currently rejects such interactions. A required coefficient that is missing, aliased (NA) or absent from the variance-covariance matrix indicates a rank-deficient design and triggers an error rather than a silently dropped term.

Value

A list with elements est, se and df.

rbmi documentation built on Oct. 15, 2026, 9:06 a.m.