Fig. 2
From: Comparison of Three Anonymization Tools for a Health Fitness Study
Normalized error for coefficients in Tables 4 and 5, and Fit (V02max predicted at mode for median distance) for Fig. 3. Normalized error E is computed as \(E=\frac{| O-A| }{\max (| O| ,| A| )}\), where O is the original value and A is the anonymized value. The middle plot displays normalized error only for the datapoints in Tables 4 and 5 where the original coefficient is significant.
