Computing Variance under Hierarchical Privacy-Related Interval Uncertainty

Formulation and Analysis of the Problem and the Resulting Algorithms

Need for hierarchical statistical analysis. In the above text, we assumed that we have all the data in one large database, and we process this large statistical database to estimate the desired statistical characteristics.

To prevent privacy violations, we replace the original values of the quasiidentifier variables with ranges. For example, we divide the set of all possible ages into ranges [0, 10], [10, 20], [20, 30], etc. Then, instead of storing the actual age of 26, we only store the range [20, 30] which contains the actual age value.