Small area estimation in R with a special focus on robustness



Release 0.2 of the rsae package has been published on May 24, 2022; see release notes: NEWS.


The rsae package is an R (R Core Team, 2022) package that provides functions to estimate the parameters of the basic unit-level model in small area estimation (also known as model type "B" in Rao, 2003, or nested-error regression model in Battese et al., 1988).

In step 1, the model is fitted by one of the methods:

  • maximum likelihood (see e.g., Rao, 2003, chapter 7.2),
  • M-estimation, which is robust against outliers; see Schoch (2012).

In step 2, the area-specific means are predicted using the empirical best linear unbiased predictor (EBLUP) or a robust prediction method due to Copt and Victoria-Feser (2009). In addition, the mean square prediction error of the area-specific means can be computed by a parametric bootstrap.


The package can be installed from CRAN using install.packages("rsae").

Code respository


  • BATTESE, G. E., R. M. HARTER, AND W. A. FULLER (1988). An error component model for prediction of county crop areas using, Journal of the American Statistical Association 83, 28–36. DOI 10.1080/01621459.1988.10478561
  • COPT, S. AND M.-P. VICTORIA-FESER (2009). Robust prediction in mixed linear models, Tech. report, University of Geneva.
  • R CORE TEAM (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL
  • RAO, J. (2003). Small Area Estimation, Hoboken (NJ): John Wiley and Sons. DOI 10.1002/0471722189
  • SCHOCH, T. (2012). Robust Unit-Level Small Area Estimation: A Fast Algorithm for Large Data, Austrian Journal of Statistics 41, 243–265. DOI 10.17713/ajs.v41i4.1548