Research map / Federated learning and privacy / Differential privacy
Comparing Population Means Under Local Differential Privacy: With Significance and Power
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Test directly on privatized bits using the transformed Bernoulli-mean null hypothesis | Transformation-based LDP test · Real-world dataset, equal-size samples | empirical type-I error: close to 0.05 | Helped |
| Estimate means and variances from two privatized bits, then plug estimates into a t-test | Estimation-based LDP test · Real-world dataset, equal-size samples | empirical type-I error: much higher type-I error than the pre-specified α | Hurt |
| Use ε = 5 local privacy instead of non-private data collection | Transformation-based LDP test · Real-world dataset, θ = 60 | sample size to achieve empirical power 0.8: roughly three times as many samples | Hurt |
| Reduce the fraction of users requiring LDP to 50% | Hybrid LDP test · Real-world dataset, θ = 60, ε = 1 | sample size to achieve high power: around half | Helped |
| Reduce the fraction of users requiring LDP to 1% | Hybrid LDP test · Real-world dataset, θ = 60, ε = 1 | sample size relative to non-private test: close to | Helped |
| Increase the mean gap θ between populations | Transformation-based LDP test · Real-world dataset, θ varied from 60 to 600 | sample size needed to achieve empirical power 0.8: θ between 60 and 600 | Helped |
| Increase ε | Transformation-based LDP test · Real-world dataset | statistical power | Helped |
Closest papers by meaning
- A Simple and Practical Method for Reducing the Disparate Impact of Differential Privacy (AAAI 2024 · similarity 0.88)
- Minimax Optimal Two-Sample Testing under Local Differential Privacy (NeurIPS 2026 · similarity 0.87)
- Analysis of Differentially Private Synthetic Data: A Measurement Error Approach (AAAI 2024 · similarity 0.85)
- Fairness Issues and Mitigations in (Differentially Private) Socio-Demographic Data Processes (AAAI 2025 · similarity 0.85)
- Finding ε and δ of Traditional Disclosure Control Systems (AAAI 2024 · similarity 0.85)
- Mean Estimation in the Add-Remove Model of Differential Privacy (ICML 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.