lineagefreq - Lineage Frequency Dynamics from Genomic Surveillance Counts
Models pathogen lineage frequency dynamics from genomic surveillance count data. Provides a unified interface for multinomial logistic regression, hierarchical partial-pooling models, and the Piantham approximation for relative reproduction number estimation. Features include rolling-origin backtesting, standardized forecast scoring, Compositional Adaptive Prediction Sets (CAPS) for horizon-aware calibrated forecasting, lineage collapsing, emergence detection, and sequencing power analysis. Designed for real-time public health surveillance of any variant-resolved pathogen. Methods described in Abousamra, Figgins, and Bedford (2024) <doi:10.1371/journal.pcbi.1012443>.
Last updated
epidemiologyforecastinggenomic-surveillancemultinomial-logistic-regressionsars-cov-2variant-frequency
5.20 score 1 stars 1 scripts 159 downloadssurvinger - Design-Adjusted Inference for Pathogen Lineage Surveillance
Provides tools for optimizing sequencing resource allocation and estimating pathogen lineage prevalence under real-world genomic surveillance conditions. Implements constrained allocation optimization for limited sequencing capacity across multiple regions and sample sources. Includes Horvitz-Thompson and post-stratified estimators that account for unequal sequencing rates, delay-adjusted nowcasting for right-censored reporting data, and combined design-weighted delay-corrected inference with uncertainty propagation.
Last updated
biostatisticsepidemiologyhorvitz-thompsonpathogen-surveillancesurvey-sampling
5.20 score 1 stars 20 scripts 496 downloadssyntheticdata - Synthetic Clinical Data Generation and Privacy-Preserving Validation
Generates synthetic clinical datasets that preserve statistical properties while reducing re-identification risk. Implements Gaussian copula simulation, bootstrap with noise injection, and Laplace noise perturbation, with built-in utility and privacy validation metrics. Useful for privacy-aware data sharing in multi-site clinical research. Validates synthetic data quality via distributional similarity (Kolmogorov-Smirnov), discriminative accuracy (real-vs-synthetic classifier), and nearest-neighbor privacy ratio. Methods described in Jordon et al. (2022) <doi:10.48550/arXiv.2205.03257> and Snoke et al. (2018) <doi:10.1111/rssa.12358>.
Last updated
clinical-datadifferential-privacyhealthcareprivacy-preservingsynthetic-data
4.30 score 1 stars 7 scripts 151 downloadsclinicalfair - Algorithmic Fairness Assessment for Clinical Prediction Models
Post-hoc fairness auditing toolkit for clinical prediction models. Unlike in-processing approaches that modify model training, this package evaluates existing models by computing group-wise fairness metrics (demographic parity, equalized odds, predictive parity, calibration disparity), visualizing disparities across protected attributes, and performing threshold-based mitigation. Supports intersectional analysis across multiple attributes and generates audit reports useful for fairness-oriented auditing in clinical AI settings. Methods described in Obermeyer et al. (2019) <doi:10.1126/science.aax2342> and Hardt, Price, and Srebro (2016) <doi:10.48550/arXiv.1610.02413>.
Last updated
algorithmic-fairnessbias-detectionclinical-aifairnesshealthcare
4.30 score 1 scripts 188 downloads