Wrapper function for as.triangle()

Description

This is a wrapper function for the ChainLadder::as.triangle() function. This is for those who have noticed column gaps when using as.triangle() to convert claims data to the triangle format.

Motivation

For example, this can occur when there are no claims in certain development periods. This affects downstream functions such as incr2cum() which may not take into account claims for some books with long tails, say, claims don’t occur for a long time e.g. between 5 and 7 development period and then claims come in in the 10th development period. This results in the cumulative triangle cutting off where you don’t expect.

This function aims to make sure that the “Triangle” input for the as.triangle() function has all the unique origin periods and all development periods are taken into account resulting in a sort of perfect half square.

This is done by creating a skeleton using the unique origin periods and sequencing the length of these origin periods from 1. The data is then merged with the skeleton then passed through the traditional as.triangle function:

as_triangle <- function(data, origin, dev, value, start = 1, delay = 1) {
  # create skeleton
  unique_origins <- unique(data[[origin]])
  dev_period <- seq(from = start, to = length(unique_origins), by = delay)
  triangle_skeleton <- expand.grid(unique_origins, dev_period, stringsAsFactors = FALSE)
  colnames(triangle_skeleton) <- c(origin, dev)

  complete_skeleton <- merge(triangle_skeleton, data[, c(origin, dev, value)], by = c(origin, dev), all.x = TRUE)
  incremental_triangle <- ChainLadder::as.triangle(
    Triangle = complete_skeleton, 
    origin = origin,
    dev = dev,
    value = value
  )

  return(incremental_triangle)
}

Limitations

I foresee a problem when you have "empty" origin periods. For example, a specific origin period between origin periods is not present in your data. The labelling of the development periods would not be accurate.