Fit a loss projection on a Triangle

Project cumulative loss across the cohort x development grid. fit_loss() is the role-specific dispatcher on the loss side – it forwards to a worker selected by method:

"ed" (default)

fit_ed() – pure exposure-driven (additive). Unconditional safe baseline; no maturity dependency.

"cl"

fit_cl() – pure Mack chain ladder (multiplicative). Classical reference.

"sa"

fit_sa() – stage-adaptive composition: ED before the maturity point, CL after.

"bf"

fit_bf() – Bornhuetter-Ferguson; requires a prior ELR (scalar or per-cohort table) passed via ....

"cc"

fit_cc() – Cape Cod (BF with a pooled ELR derived from the data).

The dispatcher returns a LossFit object whose $full schema is uniform across methods (loss_obs, loss_proj, loss_total_se, loss_ci_lo, loss_ci_hi, premium_obs, premium_proj, incr_premium_proj, plus method-specific extras). Missing slots on worker outputs (e.g. loss_ata_fit for ED, ed/selected for CL/BF/CC) are synthesized as NULL so downstream code such as fit_ratio() can guard uniformly.

Usage

fit_loss(
  x,
  method = c("ed", "cl", "sa", "bf", "cc"),
  alpha = 1,
  regime = NULL,
  premium_fit = NULL,
  premium_method = c("ed", "cl"),
  premium_alpha = 1,
  sigma_method = c("locf", "min_last2", "loglinear", "mack", "none"),
  recent = NULL,
  maturity = "auto",
  tail = FALSE,
  conf_level = 0.95,
  bootstrap = NULL,
  B = 999L,
  seed = NULL,
  type,
  ...
)

Arguments

x

A "Triangle" object. The standardized "loss" and "premium" columns are used (as_triangle() produces these).

method

One of "ed" (default), "cl", "sa", "bf", or "cc".

alpha

Variance-structure exponent for the loss fit. Default 1.

regime

Optional regime specification (loss-side). Accepts the standard 4-type dispatch (NULL / Regime / "auto" / function). Behavior depends on method: SA uses a hybrid 2-pass filter; ED / CL / BF / CC use a simple cohort cut. The same resolved regime is applied to the internal premium fit – callers needing an asymmetric loss/premium split should use fit_ratio().

premium_fit

Optional pre-built PremiumFit supplying the premium projection. Only used by "ed" (via fit_ed's internal premium handling) and "sa". When NULL, the worker calls fit_premium() internally.

premium_method

One of "ed" (default) or "cl". Used only when premium_fit = NULL for "sa".

premium_alpha

Variance-structure exponent for the premium fit. Default 1.

sigma_method

Method used to extrapolate sigma for links where it cannot be estimated. One of "locf" (default), "min_last2", "loglinear", "mack", or "none". "mack" applies the Mack (1993, Appendix B) tail estimator to the last unestimated link only, falling back to LOCF for any earlier ones with a warning. "none" performs no extrapolation; sigma stays NA and downstream variance terms drop those links via finite-value guards. Passed to .extrapolate_sigma_ata().

recent

Optional positive integer; calendar-diagonal filter.

maturity

Optional maturity specification. Accepts the standard 4-type dispatch (NULL / Maturity / "auto" / function). Only used by "cl", "sa", and "bf". Default "auto".

tail

Tail factor (logical or numeric). Forwarded to "cl" / "sa" workers. Default FALSE.

conf_level

Confidence level for analytical CI on the loss projection (loss_ci_lo, loss_ci_hi). Default 0.95.

bootstrap

Bootstrap configuration. Five forms accepted (see fit_sa() / fit_ed() / fit_cl() for method-specific defaults).

B

Integer number of bootstrap replicates. Default 999.

seed

Optional integer seed.

type

Bootstrap process type. Forwarded where applicable ("sa", "bf", "cc").

...

Method-specific arguments forwarded to the chosen worker. For method = "bf", prior is required.

Value

An object of class "LossFit". List with components: full, proj, maturity, loss_ata_fit, premium_ata_fit, premium_fit, ed, factor, selected, plus metadata.

See also

fit_ed(), fit_cl(), fit_sa(), fit_bf(), fit_cc(), fit_premium(), fit_ratio().

Examples

if (FALSE) { # \dontrun{
data(experience)
tri <- as_triangle(
  experience[coverage == "surgery"],
  groups   = "coverage",
  cohort   = "uy_m",
  calendar = "cy_m",
  loss     = "incr_loss",
  premium = "incr_premium"
)

lf    <- fit_loss(tri)                    # ED (default)
lf_cl <- fit_loss(tri, method = "cl")
lf_sa <- fit_loss(tri, method = "sa")
} # }