| Autor: |
D’Aló, Michele, Falorsi, Stefano, Fasulo, Andrea |
| Zdroj: |
METRON; 20230101, Issue: Preprints p1-15, 15p |
| Abstrakt: |
This paper describes a small area estimator based on a multivariate linear mixed model implemented in the RPackage mind. The method is a multivariate extension of the standard areaand unitsmall area estimators based on a general linear mixed model. Beyond the possibility of considering multivariate dependent variables, the proposed method allows one to consider marginal random effects, in addition to the usual area random effects. Adding marginal effects to the usual unit mixed model may help reduce the bias of small area estimates when the areas of interest are very small and/or some of them are non-sampled. The proposed multivariate method is useful every time small area methods need to be applied in a specific context, such as in the case of the Italian Permanent Census of the Population, in which one needs to estimate multiple contingency tables and estimates need to be computed for more than one particular set of unplanned domains. The proposed approach is illustrated using simulated data from the 2011 Census and it is aimed to assess the performance of EBLUP estimators defined at the unit level using the proposed formulation. |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
Full text from SpringerLink