Summarise development statistics (Mean, Median, Weighted)

S3 method for summary() on Triangle objects. Computes group-wise summary statistics for cumulative loss ratios (ratio) and per-period loss ratios (incr_ratio).

The function aggregates data by the grouping variables stored in attr(x, "groups") and the development variable stored in attr(x, "dev").

The following statistics are computed:

arithmetic mean,
median,
weighted mean (portfolio-level ratio based on sums).

Usage

# S3 method for class 'Triangle'
summary(object, ...)

Arguments

object: An object of class Triangle.
...: Unused; included for S3 compatibility.

Value

A data.table grouped by groups and dev, containing:

n_cohorts: Number of observations in the cell
ratio_mean: Mean of cumulative loss ratios
ratio_median: Median of cumulative loss ratios
ratio_wt: Weighted cumulative loss ratio (sum(loss) / sum(premium))
incr_ratio_mean: Mean of per-period loss ratios
incr_ratio_median: Median of per-period loss ratios
incr_ratio_wt: Weighted per-period loss ratio (sum(incr_loss) / sum(incr_premium))

The returned object keeps the attributes groups and dev, and its class is updated to "TriangleSummary".

Details

The weighted mean is computed as:

ratio_wt = sum(loss) / sum(premium)
incr_ratio_wt = sum(incr_loss) / sum(incr_premium)

These correspond to portfolio-level loss ratios based on premium and are typically more stable than simple averages when premium sizes differ across cohorts.

It is assumed that the input Triangle object does not contain missing values.

Examples

if (FALSE) { # \dontrun{
d <- as_triangle(
  df,
  groups   = "coverage",
  cohort   = "uy_m",
  calendar = "cy_m",
  loss     = "incr_loss",
  premium  = "incr_premium"
)
smr <- summary(d)
head(smr)
attr(smr, "longer")
} # }