Paper: “Robust estimation for small domains in business surveys: a comparison of transformation-based unit level models”

Authors: Chiara Bocci, Paul A. Smith

Contacts: Chiara Bocci <chiara.bocci@unifi.it>, Paul Smith <P.A.Smith@soton.ac.uk>
 
 
 

The dataset is described in detail in the paper (Section 3.1), but can't be freely distributed.

In order to understand the R codes, we provide the dataset columns names and the sampling procedure.

Dataset columns:
"y" - response variable (turnover 2020)
"y_2" - turnover 2018
"group" - industrial classification
"wp" - number of working persons in 2020
"sc" - size class in 2020 - Categorical variable (sc = 1, 2, 3, 4, 5)

    
"1_Dataset_Samples.r" is the program to select 1000 samples, selected following the SBS stratified design (as described in Section 3.2)

"2_HT_GREG_EBLUP.r" is the program to compute the HORVITZ-THOMPSON and GREG direct estimators (equations 1 and 2 in the paper) and the EBLUP estimator (equation 4 in the paper)

"3_EBP_PEBP.r" is the program to compute the EBP estimators (Section 2.2.1 in the paper) and Pseudo-EBP estimators (Section 2.3.1 in the paper)

"4_Karlb_MBDE_EB_SWEE.r" is the program to compute the Karlberg-type synthetic estimator (equation 6 in the paper), the Model based direct estimator (equation 8 in the paper), the EB and EB bias corrected estimators (equations 10 and 11 in the paper), and the SWEE estimator (equation 19 in the paper)

"5_summary_TotalsEstim.r" is the program to summarize the simulation results for all estimators