Register a custom modeling engine — register

Allows third-party packages to register custom engines with fit_model(). This enables an extensible plugin architecture similar to 'parsnip' engine registration.

Usage

register_engine(
  name,
  fit_fn,
  description = "",
  type = "frequentist",
  time_varying = FALSE
)

Arguments

name: Engine name (character scalar).
fit_fn: Function with signature function(data, pivot, ci_level, ...). Must return a list compatible with lfq_fit structure.
description: One-line description of the engine.
type: "frequentist" or "bayesian".
time_varying: Logical; does the engine support time-varying growth advantages?

Value

Invisibly returns the updated engine registry.

Examples

# Register a custom engine
my_engine <- function(data, pivot = NULL, ci_level = 0.95, ...) {
  # Custom implementation...
  .engine_mlr(data, pivot = pivot, ci_level = ci_level, ...)
}
register_engine("my_mlr", my_engine, "Custom MLR wrapper")
#> Registered engine "my_mlr".
lfq_engines()
#> # A tibble: 6 × 5
#>   engine   type        time_varying available description                       
#>   <chr>    <chr>       <lgl>        <lgl>     <chr>                             
#> 1 mlr      frequentist FALSE        TRUE      Multinomial logistic regression (…
#> 2 hier_mlr frequentist FALSE        TRUE      Hierarchical MLR with empirical B…
#> 3 piantham frequentist FALSE        TRUE      Piantham Rt approximation from ML…
#> 4 fga      bayesian    FALSE        FALSE     Fixed growth advantage (Stan)     
#> 5 garw     bayesian    TRUE         FALSE     Growth advantage random walk (Sta…
#> 6 my_mlr   frequentist FALSE        TRUE      Custom MLR wrapper