Business Context
A regulatory audit is required on a credit scoring model. Strict privacy laws mandate that the original training data be deleted immediately after training. The auditor must verify the model's existence and basic type using ONLY the item store, ensuring the `copulaViewStore` action functions correctly without the underlying source data.
Data Preparation
Create temporary sensitive data for model training.
Copied!
| 1 | |
| 2 | DATA mycas.sensitive_data; |
| 3 | DO i = 1 to 50; |
| 4 | x=rand('Normal'); |
| 5 | y=rand('Normal'); |
| 6 | OUTPUT; |
| 7 | END; |
| 8 | |
| 9 | RUN; |
| 10 |
Étapes de réalisation
1
Fit the model and save to 'auditStore', then immediately DELETE the source table.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | copula.copulaFit TABLE={name='sensitive_data'} copula='GAUSSIAN' var={'x', 'y'} store={name='auditStore', replace=true}; |
| 4 | TABLE.dropTable name='sensitive_data'; |
| 5 | |
| 6 | RUN; |
| 7 |
2
Attempt to view the store using 'minimal' options to verify the model artifact persists and is readable without the data.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | copula.copulaViewStore / instore={name='auditStore'} viewOptions={minimal=true}; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The action successfully reads 'auditStore' despite the source table 'sensitive_data' being deleted. It returns a minimal set of tables (Model Definition, Fit Summary), proving the item store is a self-contained artifact suitable for audit compliance.