Business Context
A retail bank is building a new credit scoring model. Before training predictive models, the data science team needs to identify highly correlated variables (multicollinearity) among customer financial metrics (Income, Debt, Credit Score) to reduce feature redundancy. Additionally, they want to verify the internal consistency reliability of these financial indicators using Cronbach's Alpha.
Data Preparation
Simulate a dataset of 5,000 banking customers with correlated variables (Income, Debt, Credit Score) and Age.
Copied!
| 1 | |
| 2 | DATA mycas.credit_risk; |
| 3 | call streaminit(123); |
| 4 | DO i=1 to 5000; |
| 5 | Income = rand('lognormal', 10, 0.5); |
| 6 | Debt = (Income * 0.4) + rand('normal', 0, 2000); |
| 7 | Age = rand('integer', 18, 75); |
| 8 | CreditScore = 800 - (Debt/100) + (Age * 1.5) + rand('normal', 0, 30); |
| 9 | OUTPUT; |
| 10 | END; |
| 11 | |
| 12 | RUN; |
| 13 |
Étapes de réalisation
1
Load the simulated data into memory (implicitly done by data step) and verify table existence.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | TABLE.tableInfo TABLE="credit_risk"; |
| 4 | |
| 5 | RUN; |
| 6 |
2
Calculate the correlation matrix including Cronbach's Alpha and save the results to a table named 'corr_stats'.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | SIMPLE.correlation TABLE={name='credit_risk'} inputs={'Income', 'Debt', 'Age', 'CreditScore'} alpha=true casOut={name='corr_stats', replace=true} descriptiveStats=true; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The action successfully computes the correlation matrix showing strong positive correlation between Income and Debt, and correlations between CreditScore and other factors. Cronbach's Alpha is calculated and displayed in the results. A new CAS table 'corr_stats' is created containing the statistical output.