Mask the last N calendar diagonals from a Triangle

Drops the most recent holdout calendar diagonals (per group) from a Triangle, returning a new Triangle of the same class with all attributes preserved. Useful for simulating a historical analyst's view – the same masking backtest() and detect_regime(holdout=) apply internally.

The calendar diagonal index is built as rank(cohort) + dev - 1, with rank() computed within group. The holdout most recent calendar indices are dropped.

Usage

mask_triangle(x, holdout = 0L)

Arguments

x

A Triangle object.

holdout

Non-negative integer. Number of latest calendar diagonals to mask. 0L (default) returns a copy of x unchanged.

Value

A Triangle with the held-out cells removed.

Examples

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

# Inspect what the analyst at a 6-month historical cutoff would see
tri_masked <- mask_triangle(tri, holdout = 6L)
plot_triangle(tri_masked)

# Use same masked tri to detect regime + fit
r   <- detect_regime(tri_masked)
fit <- fit_ratio(tri_masked, loss_regime = r)
} # }