Attribute data record quality-control results as counts or proportions, such as defective versus nondefective items, rather than continuous measurements. Their statistical treatment supports proportion-based process monitoring. section 14.1.3 section 14.2.1
For p-chart analysis, calculate and interpret attribute-data variation as follows:
- Calculate each sample proportion as defective items divided by inspected items. section 14.2.1
- Estimate standard deviation with √(p̄(1 − p̄) / n), because proportion variability depends on the average defect rate and sample size. section 14.1.3
- Set three-sigma upper and lower control limits around p̄, then inspect points for unusual variation or process instability. section 14.2.1
Orange-juice can batches
A pandas data frame contains 64 batches, each recording defective cans and batch size. Compute the defective proportion, mean proportion, sample-specific sigma, and control limits with pandas and NumPy, then plot the p-chart with Matplotlib. Points inside the limits indicate expected variation; points outside warrant investigation. section 14.2.1
