Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
MRMR Triangle constructor
# User-friendly constructor
# This is a giant pile of code which basically does the following:
# * Ensure that we have a proper date for the loss period start
# * Ensure that we have a column for the development lag
# * Once those have been established, create columns for loss period start and end, create a column
# for development period (based on lubridate Period class), compute the evaluation date.
Triangle = function(TriangleData
, TriangleName
, LossPeriodType = "accident"
, LossPeriodInterval = years(1)
, DevelopmentInterval = years(1)
, LossPeriodColumn
, DevelopmentColumn)
# confirm that the loss period column exists. If not, we throw an error.
if (!ColumnExists(TriangleData, LossPeriodColumn))
stop ("The specified column for the loss period does not exist. Unable to create the triangle.")
# if the loss period column is not a date, attempt to convert it.
if (!is.POSIXt(TriangleData[,LossPeriodColumn])){
attemptConvert = ymd(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
attemptConvert = ydm(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
attemptConvert = mdy(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
attemptConvert = myd(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
attemptConvert = dmy(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
attemptConvert = dym(TriangleData[,LossPeriodColumn])
if (sum(is.Date(attemptConvert)) ==0 ){
stop ("Loss period column is not in date form and all values cannot be converted to a date. Unable to create the triangle")
# If we've made it this far, we can safely convert the loss period start date
if (!is.POSIXt(TriangleData[, LossPeriodColumn])){
TriangleData$LossPeriodStart = attemptConvert
} else {
TriangleData$LossPeriodStart = TriangleData[, LossPeriodColumn]
# Now add the period to create the end date
TriangleData$LossPeriodEnd = TriangleData$LossPeriodStart + LossPeriodInterval - days(1)
# it's possible that the user has fed data with overlap.
# Annual data with two start dates in the same year and annual development period, zB.
# I might decide to check for this and throw a warning, but for now, I'll just blame the user.
# Now confirm that the development lag column exists. If not, we throw an error.
if (!ColumnExists(TriangleData, DevelopmentColumn))
stop ("The specified column for the development does not exist. Unable to create the triangle.")
# For now, development periods must be integer.
if (!is.integer(TriangleData[,DevelopmentColumn])){
TriangleData[,DevelopmentColumn] = as.integer(TriangleData[,DevelopmentColumn])
TriangleData$Development = TriangleData[,DevelopmentColumn] * DevelopmentInterval
# if we can, construct an evaluation date based on loss period start and development interval
TriangleData$EvaluationDate = TriangleData$LossPeriodStart + TriangleData$Development - days(1)
tri = new("Triangle", TriangleData = TriangleData
, TriangleName = TriangleName
, LossPeriodType = LossPeriodType
, LossPeriodInterval = LossPeriodInterval
, DevelopmentInterval = DevelopmentInterval)
tri@TriangleData = CalculateIncrementals(tri)
return (tri)
#== End Triangle constructor ======================================================================
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment