FIMA builds differentially private confidence intervals and hypothesis tests for categorical data by matching the released statistic to simulated noisy versions via a fiducial solution.
To evaluate consistency, we repeated the private test 104 times with different random seeds, observing a minimum p-value of 0 and a maximum of 0.0062
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ME 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fiducial Matching: Differentially Private Inference for Categorical Data
FIMA builds differentially private confidence intervals and hypothesis tests for categorical data by matching the released statistic to simulated noisy versions via a fiducial solution.