Learn R Programming

nlfh (version 0.1.0)

nlfh-package: nlfh: Nonlinear Fay-Herriot Models for Small Area Estimation

Description

nlfh fits Bayesian Fay-Herriot models for area-level small area estimation. The package works with direct estimates, known sampling variances, and area-level covariates, returning posterior draws and summaries for latent area quantities and model parameters.

Arguments

Author

Maintainer: Paul Parker paulparker@ucsc.edu

Details

The main entry point is fit_fh(). It supports a formula interface for specifying the response and covariates, and a matrix interface for workflows that already construct design matrices. Fitted model objects provide compact printing, posterior summaries via summary(), fitted area estimates via fitted(), and retained MCMC draws via posterior_draws().

Available model families are:

  • method = "linear": the standard linear Fay-Herriot model.

  • method = "rnn": a random-weight neural network Fay-Herriot extension for nonlinear mean functions. Non-intercept covariates are centered and scaled by default, and the response is standardized internally before returned quantities are transformed back to the original scale.

  • method = "bart": a BART Fay-Herriot model for flexible nonlinear and interaction effects. The first model-matrix column is treated as a baseline/intercept column and is excluded from BART variable importance.

See Also