Algorithm 1: Private and robust estimator for (m_2) | Algorithm 2: Unbounded DP upper-quantile estimator | Algorithm 3: Additive DP histogram | Algorithm 4: Sparse DP histogram | Algorithm 5: Group-wise additive DP histogram | Algorithm 6: Group-wise sparse DP histogram | References
Package: dppca 0.2.0
dppca: Differentially Private Principal Component Analysis and Visualization
Provides tools for differentially private principal component analysis (PCA) and visualization. Includes private principal component directions, scree and proportion of variance explained (PVE) estimates, and two-dimensional private score histograms with synthetic sampled points. Supports additive and sparse histogram mechanisms, group-wise score visualizations, and optional non-private comparisons. An integrated workflow produces loading plots, scree plots, score plots, variable plots, and biplots from shared private PCA estimates. An interactive 'shiny' application is also provided. Private direction estimation is based on Kim and Jung (2025) <doi:10.1002/sam.70053>. Private scree estimation draws on Dwork and Roth (2014) <doi:10.1561/0400000042>, Ramsay and Spicker (2025) <doi:10.48550/arXiv.2501.14095>, and Yu, Ren and Zhou (2024) <doi:10.3150/23-BEJ1706>. Private score frames use unbounded quantile mechanisms from Durfee (2023) <doi:10.52202/075280-3394>. Private score histograms use methods related to Wasserman and Zhou (2010) <doi:10.1198/jasa.2009.tm08651> and Karwa and Vadhan (2018) <doi:10.4230/LIPIcs.ITCS.2018.44>.
Authors:
dppca_0.2.0.tar.gz
dppca_0.2.0.zip(r-4.7-x86_64)dppca_0.2.0.zip(r-4.7-arm64)dppca_0.2.0.zip(r-4.6-x86_64)dppca_0.2.0.zip(r-4.6-arm64)dppca_0.2.0.zip(r-4.5-x86_64)
dppca_0.2.0.tgz(r-4.6-x86_64)dppca_0.2.0.tgz(r-4.6-arm64)dppca_0.2.0.tgz(r-4.5-x86_64)dppca_0.2.0.tgz(r-4.5-arm64)
dppca_0.2.0.tar.gz(r-4.7-arm64)dppca_0.2.0.tar.gz(r-4.7-x86_64)dppca_0.2.0.tar.gz(r-4.6-arm64)dppca_0.2.0.tar.gz(r-4.6-x86_64)
dppca_0.2.0.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION
card.svg |card.png
dppca/json (API)
| # Install 'dppca' in R: |
| install.packages('dppca', repos = c('https://yejinjo0220.r-universe.dev', 'https://cloud.r-project.org')) |
Bug tracker:https://github.com/yejinjo0220/dppca/issues
Pkgdown/docs site:https://yejinjo0220.github.io
Datasets:
Last updated from:a924151208. Checks:15 OK. Indexed: yes.
| Target | Result | Time | Files | Syslog |
|---|---|---|---|---|
| linux-devel-arm64 | OK | 135 | ||
| linux-devel-x86_64 | OK | 180 | ||
| source / vignettes | OK | 370 | ||
| linux-release-arm64 | OK | 132 | ||
| linux-release-x86_64 | OK | 167 | ||
| macos-release-arm64 | OK | 100 | ||
| macos-release-x86_64 | OK | 373 | ||
| macos-oldrel-arm64 | OK | 127 | ||
| macos-oldrel-x86_64 | OK | 246 | ||
| windows-devel-arm64 | OK | 226 | ||
| windows-devel-x86_64 | OK | 151 | ||
| windows-release-arm64 | OK | 227 | ||
| windows-release-x86_64 | OK | 123 | ||
| windows-oldrel-x86_64 | OK | 116 | ||
| wasm-release | OK | 107 |
Exports:clipped_controldp_loadingdp_pcadp_scoredp_screedppca_apphuber_controlloading_plot_controlpmwm_controlprivacy_controlsampling_controlscore_plot_controlscree_plot_control
Dependencies:clicpp11farverggplot2ggrepelgluegtableisobandlabelinglifecyclepatchworkR6rbibutilsRColorBrewerRcppRdpackrlangS7scalesvctrsVGAMviridisLitewithr
Algorithms
Last update: 2026-10-28
Started: 2026-05-05
Differentially Private Eigenvalues
1. DP eigenvalue estimation | $$(\widetilde{\lambda}_1,\ldots,\widetilde{\lambda}_k) | 2. Clipped mean | 3. Private modified winsorized mean (PMWM) | 3.1. Allocation of pair rows | 3.2. Private quantiles and cutoff bounds | $$\zeta | $$L_\ell^ | \mathcal Q_ | $$L_\ell | \min{b,\max{a,L_\ell^{\mathrm{priv}}}},\qquadU_\ell | 3.3. Winsorized mean and Gaussian perturbation | 3.4. Privacy allocation and tuning choices | $$(\varepsilon_M,\delta_M) | 4. Huber mean | 4.1. Huber loss and score | 4.2. Private scale and Huber thresholds | $$\varepsilon_ | \frac | $$\tau_\ell | 4.3. Noisy gradient descent | 5. Nonnegative scree and PVE | 6. DP scree/PVE plots in dppca | References
Last update: 2026-10-28
Started: 2026-10-28
Differentially Private PC Directions
1. Spherical Kendall matrix and PC directions | Translation invariance | 2. DP PC directions | 3. DP loading plots in dppca | References
Last update: 2026-10-28
Started: 2026-10-28
Differentially Private PC Scores
1. Two-dimensional PC scores | 2. Private plotting frame | 2.1. Private coordinate medians | 2.2. Private radial quantile and square frame | 3. Histogram grid and sensitivity | 4. Private histogram estimators | 4.1. Additive Gaussian histogram | 4.2. Stability-based sparse histogram | 5. Privacy budget allocation | 6. Synthetic score sampling | 7. Group-wise private score histograms | 8.1. Display centering | $$c_\ell^ | References
Last update: 2026-10-28
Started: 2026-10-28
Differentially Private PCA Visualizations
1. What is estimated privately? | 2. Privacy budget and shared estimation | 3. Five DP PCA Visualization | 3.1. DP variable plot (correlation-style coordinates) | $$\operatorname{Cor}(X_{\cdot j},S_{\cdot\ell}) | $$\widetilde{\Sigma}_k | $$\widetilde{\sigma}_{j,k}^2 | (\widetilde | $$\widetilde | $$R_j^ | $$|R_j^{\mathrm{DP}}|_2^2 | \widetilde | 3.2. DP biplot | $$\widetilde{d}_\ell | $$G_{i\ell}^{\mathrm{DP}}(\alpha) | \frac{s_{i\ell}^*-c_\ell^{\mathrm{disp}}}{\widetilde{d}\ell^\alpha},\qquadH{j\ell}^{\mathrm{DP}}(\alpha) | For the two selected components, the rows $G_i^{\mathrm{DP}}(\alpha) | \left(G_{ia}^{\mathrm{DP}}(\alpha),G_{ib}^{\mathrm{DP}}(\alpha)\right)$represent synthetic observations, while $H_j^{\mathrm{DP}}(\alpha) | Let$$r_G | Q_{0.9}\left(\left{\left|G_i^{\mathrm{DP}}(\alpha) | $$r_H | $$\gamma | 4. Summary | References and further reading
Last update: 2026-10-28
Started: 2026-10-28
DP Loading Plots
1. Load the example data | 2. Set the privacy budget and estimate PC directions | 3. Inspect the result | 4. Plot the result | 4.1. Vector plot | 4.2. Point and heatmap plots | 5. Select PCs and variables | 6. Customize the appearance
Last update: 2026-10-28
Started: 2026-10-27
DP PCA Visualization
What you will make | 1. Load the vowel data | 2. Create the privacy allocation | 3. Fit DP PCA analysis | Main arguments of dp_pca() | 4. Inspect the results | 5. Create all five displays | 6. Reuse and export the results
Last update: 2026-10-28
Started: 2026-10-27
DP Score Plots
1. Load the example data | 2. Set the privacy budget and estimate a score histogram | 2.1. Additive and sparse methods | 3. Inspect and plot the result | 3.1. Read the fit and stored estimates | 3.2. Show a private histogram | 3.3. Draw synthetic score points | 4. Customize the plot | 5. Group-wise private score histograms | 5.1. Fit a private histogram for each group | 5.2. Plot a group histogram overlay
Last update: 2026-10-28
Started: 2026-10-27
DP Scree Plots
1. Load the example data | 2. Set the privacy budget and estimate scree values | 3. Inspect and plot the result | 4. Choose a method and its tuning parameters | Clipped mean | Private modified winsorized mean (PMWM) | Huber mean | 5. Compare multiple methods and account for releases | 6. Reusing Private PC Directions | 7. Customize the plot
Last update: 2026-10-28
Started: 2026-10-27
Getting Started with dppca
Choose a starting point | 1. Install and load the package | 2. Prepare your input data | 3. Explore the Example Datasets | 3.1. Gaussian data | 3.2. Vowel data | 4. Set and Allocate the Privacy Budget | 4.1. Default privacy budget allocation | 4.2. Customize the budget allocation | 4.3. Specify individual component budgets | 5. Next steps
Last update: 2026-10-28
Started: 2026-10-27
Interactive Shiny App
1. Available analyses | 2. Getting started | 3. Data and comparison options | 4. Run the app locally
Last update: 2026-10-28
Started: 2026-10-27
Principal Component Analysis
1. Introduction | 2. Population PCA | Let $\xi\in\mathbb{R}^p$ be a random vector with mean $\mu=E[\xi]$ and covariance matrix $\Sigma | 3. Sample PCA | [\widehat \Sigma | 3.1 Directions | [\widehat \Sigma \widehat v_\ell | 3.2 Eigenvalues | The eigenvalues of $\widehat \Sigma$ represent the sample variances along the corresponding PC directions. In particular,[\widehat\lambda_\ell | [\operatorname{PVE}_\ell | 3.3 Scores | where the $\ell$th PC score of observation $i$ is $s_ | 4. PCA visualizations | 4.1 Loading plot | 4.2 Scree plot | 4.3 Score plot | 4.4 Variable plot (correlation circle) | A variable plot displays the correlations between the original variables and the PC scores. Let (\widehat \sigma_j=\sqrt{\widehat \Sigma_{jj}}) denote the sample standard deviation of variable (j). For (\widehat \sigma>0) and (\widehat\lambda_\ell>0), the sample correlation between variable (j) and the (\ell)th PC scores is[\operatorname{Cor}(X_{\cdot j},S_{\cdot\ell}) | Since $|R_j|_2^2 | 4.5 Biplot | Summary of PCA visualizations | References
Last update: 2026-10-28
Started: 2026-10-28
