The fitting functions in this package fall back to a simpler specification when covariate selection returns nothing or the adjusted fit is degenerate. These accessors report what was actually run, so a fallback is visible in the results rather than buried inside the wrapper.
Arguments
- fit
A fit returned by
lm_lin_lasso,lm_robust_lasso, orlm_moderator_lasso.
Value
adjustment() returns "lin", "robust", or
"none". selected_covariates() returns the character vector
of covariates actually used, possibly empty. fallback_reason()
returns NA_character_ when no fallback fired, and otherwise a
description of why it did.
See also
Other fallback reporting:
fallback_log(),
fallback_summary()
Examples
set.seed(1)
n <- 300
dat <- data.frame(Z = rep(0:1, n / 2), X1 = rnorm(n))
dat$Y <- 0.5 * dat$Z + rnorm(n)
# X1 is pure noise, so selection returns nothing and the fit falls back
fit <- lm_lin_lasso(Y ~ Z, ~ X1, data = dat)
adjustment(fit)
#> [1] "lin"
fallback_reason(fit)
#> [1] NA