Extract covariate names for longitudinal model parameters

Uses the subject-level design matrices prepared for fitting the model. Covariate names therefore include factor contrasts, transformations, and interactions, in the same order as the corresponding coefficients. Covariates are taken from subject data, rather than repeated measurements in a DataLongitudinal object.

Usage

# S3 method for class 'LongitudinalModel'
covariates(object, data, ...)

Arguments

object

(LongitudinalModel) the longitudinal model specification.

data

(DataSubject) the subject data used to construct the model's design matrices.

...

not used.

Value

A named list of character vectors, keyed by coefficient parameter name. Intercept-only predictors have an empty character vector. Models without covariate predictors, such as LongitudinalSteinFojo, return an empty list.

Details

The default longitudinal-model method recognises design matrices named <name>_design paired with declared parameters named <name>_coefficients. Custom longitudinal models using another convention can provide their own covariates() method with the same return structure.

See also

Other covariates: covariates(), covariates.DataSurvival()

Examples

subject <- DataSubject(
    data.frame(
        id = c("1", "2"),
        arm = c("A", "B"),
        study = "Study 1",
        age = c(50, 60)
    ),
    "id", "arm", "study"
)
model <- LongitudinalSteinFojoCov(mu_g_formula = ~ arm + age)
covariates(model, data = subject)$lm_sfc_mu_g_coefficients
#> [1] "armB" "age" 
covariates(LongitudinalSteinFojo(), data = subject)
#> list()