Overview | 1. Spatial Fay-Herriot Model (eblup_sfh) | Model Formulation | Fitting the Model | Unsampled Domains and Spatial Kriging | Multi-Threaded Bootstrap MSE | 2. Spatio-Temporal Fay-Herriot Model (eblup_stfh) | Fitting the Spatio-Temporal Model | Unsampled Domains and Missing Periods | 3. Comparing Models with autoplot | 4. Bayesian Spatio-Temporal Modeling (hb_area) | References
Package: fastsae 1.0.0
fastsae: High-Performance Frequentist and Bayesian Small Area Estimation in R
High-performance Small Area Estimation (SAE) in R, implementing frequentist EBLUP models via compiled C++ and Bayesian hierarchical models via 'INLA' (Integrated Nested Laplace Approximations) or, optionally, full MCMC with 'NumPyro' (Phan et al., 2019) <doi:10.48550/arXiv.1912.11554> on 'JAX' (CPU or NVIDIA GPU). Frequentist models include the area-level Fay-Herriot model (Fay and Herriot, 1979 <doi:10.1080/01621459.1979.10482505>), spatial Fay-Herriot (Pratesi and Salvati, 2008 <doi:10.1007/s10260-007-0061-9>), spatio-temporal Fay-Herriot (Marhuenda et al., 2013 <doi:10.1016/j.csda.2012.09.002>), unit-level Battese-Harter-Fuller (Battese et al., 1988 <doi:10.1080/01621459.1988.10478561>), and two-fold subarea models (Torabi and Rao, 2014 <doi:10.1016/j.jmva.2014.02.001>) for nested area-subarea hierarchies. Bayesian models use 'INLA' (Rue et al., 2009 <doi:10.1111/j.1467-9868.2008.00700.x>), supporting 6 likelihood families (Gaussian, Beta, Binomial, Poisson, Negative Binomial, Gamma), spatial structures (BYM2, Besag ICAR), temporal dynamics (RW1, RW2, AR1), and spatio-temporal interactions. Key features include analytical and bootstrap MSE estimation, benchmarking to aggregate targets (Rao and Molina, 2015 <doi:10.1002/9781118735855>), automatic handling of unsampled domains, and S3 methods for summaries, diagnostics and plots.
Authors:
fastsae_1.0.0.tar.gz
fastsae_1.0.0.zip(r-4.7-x86_64)fastsae_1.0.0.zip(r-4.7-arm64)fastsae_1.0.0.zip(r-4.6-x86_64)fastsae_1.0.0.zip(r-4.6-arm64)fastsae_1.0.0.zip(r-4.5-x86_64)
fastsae_1.0.0.tgz(r-4.6-x86_64)fastsae_1.0.0.tgz(r-4.6-arm64)fastsae_1.0.0.tgz(r-4.5-x86_64)fastsae_1.0.0.tgz(r-4.5-arm64)
fastsae_1.0.0.tar.gz(r-4.7-arm64)fastsae_1.0.0.tar.gz(r-4.7-x86_64)fastsae_1.0.0.tar.gz(r-4.6-arm64)fastsae_1.0.0.tar.gz(r-4.6-x86_64)
fastsae_1.0.0.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION |NEWS
card.svg |card.png
fastsae/json (API)
| # Install 'fastsae' in R: |
| install.packages('fastsae', repos = c('https://ridsonap.r-universe.dev', 'https://cloud.r-project.org')) |
Bug tracker:https://github.com/ridsonap/fastsae/issues
Pkgdown/docs site:https://ridsonap.github.io
Uses libs:
Datasets:
- cornsoybean - Corn and Soybean Survey and Satellite Data in 12 Iowa Counties
- cornsoybeanmeans - Corn and Soybean Mean Number of Pixels per Segment for 12 Iowa Counties
- mys - Mys: mean years of schooling people with disabilities.
- mys_panel - Mys_panel: mean years of schooling people with disabilities 2016 - 2026.
- mys_proxmat - Example proximity matrix
- sim_area - Multi-Distribution Synthetic Area-Level Dataset
- sim_panel - Multi-Distribution Spatio-Temporal Panel Dataset
cppinlaopenmpsmall-area-estimationopenblascpp
Last updated from:7e20210fc3. Checks:4 ERROR, 11 OK. Indexed: yes.
| Target | Result | Time | Files | Syslog |
|---|---|---|---|---|
| linux-devel-arm64 | ERROR | 326 | ||
| linux-devel-x86_64 | OK | 592 | ||
| source / vignettes | OK | 637 | ||
| linux-release-arm64 | ERROR | 403 | ||
| linux-release-x86_64 | OK | 639 | ||
| macos-release-arm64 | OK | 948 | ||
| macos-release-x86_64 | OK | 1566 | ||
| macos-oldrel-arm64 | ERROR | 320 | ||
| macos-oldrel-x86_64 | ERROR | 759 | ||
| windows-devel-arm64 | OK | 498 | ||
| windows-devel-x86_64 | OK | 684 | ||
| windows-release-arm64 | OK | 446 | ||
| windows-release-x86_64 | OK | 505 | ||
| windows-oldrel-x86_64 | OK | 891 | ||
| wasm-release | OK | 362 |
Exports:add_reliability_flagsautoplotbenchmark_saecheck_numpyro_availablecompare_saecreate_weightsdiagnoseeblup_bhfeblup_fheblup_sfheblup_stfheblup_twofoldexport_saegvf_smoothhb_areahb_twofoldhb_unitinla_enginejax_enginemap_saesetup_numpyro_envsim_area_datasim_series_datasim_spatial_weightsspatial_test
Dependencies:clicpp11farverggplot2gluegtableisobandlabelinglatticelifecycleMatrixR6RColorBrewerRcppRcppArmadilloRcppEigenrlangS7scalesvctrsviridisLitewithr
Spatial and Spatio-Temporal Models
Last update: 2026-10-14
Started: 2026-09-21
Model Diagnostics and Residual Analysis
Introduction | Core Diagnostic Criteria | Practical Example: Diagnosing a Fay-Herriot Model | 1. Fit Model and Run diagnose() | 2. Inspect Diagnostic Components | Visual Diagnostics with autoplot() | 1. Calibration Plot (type = "calibration") | 2. Precision Comparison (type = "rse") | 3. Residual Diagnostics (type = "residuals" & type = "qq") | 4. Combined Diagnostic Dashboard (type = "all") | Diagnosing Bayesian Spatio-Temporal Models | References
Last update: 2026-10-13
Started: 2026-09-26
Bayesian Area-Level Small Area Estimation with INLA
Introduction | Supported Distribution Families | Mathematical Formulation | Level 1: Sampling Model (Likelihood) | Level 2: Prior on Fixed Effects | Level 3: Prior on Random Effects | INLA Approximation | Spatial and Spatio-Temporal Structures | Spatial Priors | Temporal Models | Space-Time Interaction Structures (st_interaction) | Case Study: Spatio-Temporal Beta SAE on Emilia-Romagna Poverty | 1. Load Data and Construct Spatial Adjacency Matrix | 2. Fit Spatio-Temporal Beta Model | 3. Inspect Posterior Predictions and Uncertainty | Comparison and Validation Against tipsae (Stan MCMC) | Interactive Scalability Explorer ($n = 30$ to $n = 1,000$) | Inferential Equivalence & Validation Metrics ($n = 1,000$) | Cross-Sectional Count Data: Poisson & Negative Binomial | References
Last update: 2026-10-13
Started: 2026-09-29
Performance & Scalability Benchmarks
Executive Summary | Benchmark Results Table | Interactive Benchmark Explorer | Architectural Insights: Why is fastsae so Fast? | Bayesian Spatio-Temporal Benchmark: fastsae (INLA) vs tipsae (Stan MCMC) | Interactive Beta SAE Benchmark Explorer ($n = 30$ to $n = 1,000$) | Inferential Equivalence & Validation Metrics ($n = 1,000$) | Finite Population Simulation & Ground Truth Validation
Last update: 2026-10-13
Started: 2026-09-21
Finite Population Simulation: Validating fastsae against Ground Truth
Introduction: Why Finite Population Simulation? | Simulation Architecture | Mathematical Formulations of Empirical Properties | Empirical Benchmark Results ($R = 100$ Replications, $D = 60$) | Key Empirical Findings: | National-Scale Benchmark ($D = 500$ Domains, $N \approx 250{,}000$ Individuals, $R = 25$) | Key National Scalability Takeaways: | Visualizations | 1. Domain-Specific Empirical RMSE | 2. Empirical 95% Credible Interval Coverage | How to Replicate | References
Last update: 2026-10-13
Started: 2026-09-27
Getting Started with fastsae
Introduction | The Fay-Herriot Model | Step-by-Step Example | 1. Load Package and Dataset | 2. Fit Fay-Herriot Model (eblup_fh) | 3. Model Summary and Coefficients | 4. Diagnostic Plots (autoplot) | Direct Estimates vs EBLUP | Mean Squared Error (MSE) Across Domains | Exact Numerical Equivalence with sae | Next Steps | References
Last update: 2026-09-29
Started: 2026-09-21
Unit-Level Estimation with Battese-Harter-Fuller
Overview | Model Formulation | Step-by-Step Example | 1. Data Preparation | 2. Fit BHF Model with Bootstrap MSE | 3. Inspect Domain Estimates | 4. Visualizing Results with autoplot | References
Last update: 2026-09-21
Started: 2026-09-21
