Loss reserving has a new, silly name
January 18, 2013 2 Comments
I started using Git some time ago, but mostly for local work files. Today, I finally sync’ed up a repository for loss reserving analysis. It may be found here: https://github.com/PirateGrunt/MRMR
MRMR stands for Multivariate Regression Model for Reserves. When pronounced “Mister Mister” it also sounds like a thankfully forgotten American soft pop band from the ’80s (“Kyrie”, “Broken Wings”, etc.). It may also bring to mind MCMC, a subject that I’m trying to master in my spare time.
Current capabilities are:
- Fetch NAIC data from the CAS research site.
- Prepare a design matrix to fit a linear model.
- Fit a weighted OLS regression to a loss triangle.
- Write some basic diagnostic graphs to assess quality of the model.
This was some of the first R code that I wrote about a year ago and it’s been a while since I’ve looked at it. I expect I’ll be making loads of changes over the next few weeks. UPDATE: I tried to run the projection code this afternoon and got some incorrect results. At present, the projection function is pants. I hope to have it sorted out by the end of the weekend.
If you’d like to contribute, please let me know. I couldn’t possibly compete with ChainLadder. This is more just for fun and a convenient way to store and share code.
Here’s the demo script and a picture of the output.
source("https://raw.github.com/PirateGrunt/MRMR/master/RegressionSupport.r") source("https://raw.github.com/PirateGrunt/MRMR/master/NAIC.R") source("https://raw.github.com/PirateGrunt/MRMR/master/ReservingVisualization.R") source("https://raw.github.com/PirateGrunt/MRMR/master/Triangle.R") df = GetNAICData("wkcomp_pos.csv") bigCompany = as.character(df[which(df$CumulativePaid == max(df$CumulativePaid)),"GroupName"]) df.BigCo = subset(df, GroupName == bigCompany) df.UpperTriangle = subset(df.BigCo, DevelopmentYear <=1997) tri = Triangle(TriangleData = df.UpperTriangle , TriangleName = bigCompany , LossPeriodType = "accident" , LossPeriodInterval = years(1) , DevelopmentInterval = years(1) , LossPeriodColumn = "LossPeriodStart" , DevelopmentColumn = "DevelopmentLag") tri@TriangleName tri is(tri, "Triangle") is.Triangle(tri) plt = ShowTriangle(tri@TriangleData, bigCompany) plot(tri) head(LatestDiagonal(tri)) length(LatestDiagonal(tri)[,1]) plt = ShowTriangle(tri@TriangleData, bigCompany, Cumulative=FALSE) #Note the apparent calendar year impact in 1996. This is invisible in the cumulative display.