Business Context
A banking institution needs to perform a stress test on its operational risk models. They require a high volume of simulations (100,000 replicates) to ensure tail accuracy and want to assess the stability of the model by perturbing the parameters of the severity distribution to analyze sensitivity.
Data Preparation
Creation of a severity definition table for a heavy-tailed Burr distribution, often used in operational risk.
Copied!
| 1 | |
| 2 | DATA mycas.sev_def_risk; |
| 3 | LENGTH model $8; |
| 4 | model='Burr'; |
| 5 | alpha=2; |
| 6 | theta=5000; |
| 7 | gamma=0.8; |
| 8 | dist='BURR'; |
| 9 | OUTPUT; |
| 10 | |
| 11 | RUN; |
| 12 |
Étapes de réalisation
1
Running CDM with high replication count and parameter perturbation enabled (ignoring covariance for this test case).
Copied!
| 1 | PROC CAS; |
| 2 | cdm.cdm / |
| 3 | seed=999, |
| 4 | nReplicates=100000, |
| 5 | aggLossModelType='COLLECTIVERISK', |
| 6 | countDistributions={{name='POISSON', parmValues={{value=50}}}}, |
| 7 | severityDefinitions={name='mycas.sev_def_risk'}, |
| 8 | nPerturbedSamples=10, |
| 9 | ignoreParmCovariance=TRUE, |
| 10 | OUTPUT={outSample={name='risk_stress_sim', replace=true}, perturbOut=true}; |
| 11 | RUN; QUIT; |
Expected Result
The system should handle the high volume of 100,000 replicates without memory errors. The output table 'risk_stress_sim' must contain the main simulation plus 10 additional sets of results corresponding to the perturbed parameters, allowing the bank to analyze the variance in risk estimates.