Business Context
A pharmaceutical company is conducting a clinical trial analysis. The dataset contains patient vitals and drug response rates. However, the data is 'dirty': some patients dropped out (missing values), and the study design requires weighting patients based on demographic representation (Sample Weights). The analysis must strictly exclude any incomplete records to meet regulatory standards.
Data Preparation
Create a dataset with missing values (nulls) and a weighting variable. 'Response' is missing for 10% of records.
Copied!
| 1 | |
| 2 | DATA mycas.clinical_study; |
| 3 | call streaminit(456); |
| 4 | DO i=1 to 200; |
| 5 | Demog_Weight = rand('uniform', 0.5, 1.5); |
| 6 | Dosage = rand('integer', 50, 200); |
| 7 | Response = (Dosage * 0.8) + rand('normal', 0, 10); |
| 8 | IF rand('uniform') < 0.1 THEN Response = .; |
| 9 | OUTPUT; |
| 10 | END; |
| 11 | |
| 12 | RUN; |
| 13 |
Étapes de réalisation
1
Run the correlation analysis using 'Demog_Weight' for weighting, applying listwise deletion for missing values, and using the sum of weights (WEIGHT) as the variance divisor.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | SIMPLE.correlation TABLE={name='clinical_study'} inputs={'Dosage', 'Response'} weight='Demog_Weight' listwiseDelMiss=true varianceDivisor='WEIGHT'; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The output correlation matrix is calculated using the specified weights. The number of observations used is less than the total input rows (200), confirming that rows with missing 'Response' values were successfully excluded (listwise deletion). The variance/covariance calculations utilize the sum of weights divisor as requested.