Meyers Flashcards
Three tests for uniformity for n predicted percentiles
Histogram
p-p plot
K-S test
Mack model performance on incurred and paid loss data
Light tails on incurred
Biased high on paid
ODP model performance on paid loss data
Biased high
Two reasons Mack model does not validate against paid and/or incurred
Insurance loss environment has experienced changes not yet observable
Other models that can be validated
S-shape in Meyers
Light-tailed
Reverse S-shape in Meyers
Heavy-tailed
U-shape in Meyers
Biased high
Two ways to increase variability of predictive distribution produced by Mack model on incurred
1) Treat level of accident year as random instead of fixed (LCL)
2) Allow for correlation between AYs (CCL)
LCL performance on incurred data
Increased variability from Mack, produced light tails; failed KS
CCL performance on incurred data
Increased variability from LCL; validated
Two consequences of including payment year trend
Model should be based on incremental pad loss rather than cumulative
Incremental paid loss tend to be skewed to right and occasionally negative
Two formulation for skew normal
Mixed truncated normal-normal
Mixed lognormal-normal
Two Bayesian models that include payment year trend used to model paid losses
CIT (allows for correlation between AYs)
LIT (does not allow for correlation)
LIT performance on paid data
Biased high
CIT performance on paid data
Biased high
