Business Context
An insurance actuarial team needs to perform a 'what-if' analysis on claim severity frequencies. Instead of defining parameters manually, they must reuse a specific Clayton copula model that was fitted and approved in a previous regulatory audit. This ensures consistency with the approved model parameters.
Data Preparation
Creation of base claim data to fit the initial model that will be stored.
Copied!
| 1 | DATA claims_data; |
| 2 | DO i=1 to 1000; |
| 3 | Claim_A = rand('EXPONENTIAL'); |
| 4 | Claim_B = Claim_A + rand('UNIFORM'); |
| 5 | OUTPUT; |
| 6 | END; |
| 7 | RUN; |
| 8 | |
| 9 | PROC CASUTIL; |
| 10 | load DATA=claims_data casout='claims_data' replace; |
| 11 | QUIT; |
Étapes de réalisation
1
Fit a Clayton copula and save the model definition to an item store.
Copied!
| 1 | PROC CAS; |
| 2 | copula.copulaFit / |
| 3 | TABLE={name='claims_data'}, |
| 4 | define={copulaType='CLAYTON', var={'Claim_A', 'Claim_B'}}, |
| 5 | store={name='approved_model_store', replace=true}; |
| 6 | RUN; |
| 7 | QUIT; |
2
Simulate new data by restoring the previously fitted model from the item store.
Copied!
| 1 | PROC CAS; |
| 2 | copula.copulaSimulate / |
| 3 | restore={name='approved_model_store'}, |
| 4 | ndraws=5000, |
| 5 | seed=67890, |
| 6 | outuniform={name='forecasted_claims', replace=true}; |
| 7 | RUN; |
| 8 | QUIT; |
Expected Result
First, an item store 'approved_model_store' is successfully created containing the fitted Clayton parameters. Second, the simulation produces 'forecasted_claims' (5,000 rows) strictly adhering to the dependency structure retrieved from the stored model.