CEMA SAE Short Course
Course summary for CEMA SAE Short Course
Small area estimation is of crucial importance in low- and middle-income countries (LMICs). A modern Bayesian treatment will be presented and illustrated using a range of examples. Area-level (Fay–Herriot) and unit-level models will be presented. Unit-level models for both linear and generalized linear models will be discussed. Fast computation is carried out with the Integrated Nested Laplace Approximation (INLA) method, which is embedded within the SUMMER and surveyPrev R packages. Hyperprior specification is via penalized complexity priors. Between-area variation will be modeled using independent and spatial random effects. For the latter, the Besag, York, Mollié model will be described.
Short Course Slides
- Overview & Schedule / Context and Motivation: Slides
- Pre-modeled app: Slides
- Survey sampling: Slides
- Weighted estimation/summarizing uncertainty: Slides
- Weighted estimation in the DHS app: Slides
- Bayesian smoothing: Slides
- Fay-Herriot models, Basic: Slides
- Fay-Herriot models, Advanced: Slides
- Use of AI in software development: Slides
- Fay-Herriot covariate modeling with the DHS app: Slides
- Unit-level models: Slides
- Overview of computing: Slides
- Building the apps: Slides
- Model extensions: Slides
More to be uploaded.
Covariate Datasets
- Kenya County Covariate - Elevation: Data
- Kenya County Covariate - ITN Access: Data
- Kenya County Covariate - Lowest Wealth Quintile: Data
- Kenya County Covariate - NDVI: Data
- Kenya County Covariate - Precipitation: Data
- Kenya County Covariate - Temperature: Data
- Kenya Sub-County Covariate - Average Education Years: Data
R Packages:
surveyPrev with vignette on prevalence mapping and vignette on creating indicators
SUMMER See the cran site for various vignettes
DHS Data:
Demographic and Heath Surveys (DHS) data can be downloaded, after registering for an account here
When requesting specific datasets, remember to request the GPS data (locations of clusters)
Web Apps
MICS R Shiny App (beta test version)
Papers: