Calibration methods for small domain estimation

Risto Lehtonen, Ari Veijanen

Research output: Contribution to journalArticleScientific

Abstract

Abstract. Small area estimation was discussed in this section of The Survey Statistician in the July 2010 issue by Danny Pfeffermann in his article "Small Area Estimation: Basic Concepts, Models and Ongoing Research". A more comprehensive review article was published a couple years later (Pfeffermann 2013). The Wiley book "Small Area Estimation, Second Edition" of 2015 by J.N.K. Rao and Isabel Molina presents an update to Rao's monograph of 2003 on small area estimation. These important sources cover model-based and design-based approaches on small area estimation (SAE) and show in particular the progress in model-based methods, and the progress is ongoing. In this article we introduce methods that incorporate assisting models in a design-based estimation procedure for population characteristics (totals, means etc.) for subgroups or domains, including small domains (with small sample size). We use logistic mixed models in model-assisted calibration estimation of poverty rates for administrative regions (domains of interest). Statistical properties (design bias and accuracy) of the method is compared with the classical model-free calibration method of Deville and Särndal (1992) and further, with a model-based SAE method that relies on the same logistic mixed model as the model-assisted counterpart. Our design-based simulation experiments employ real data obtained from registers of Statistics Finland. The paper is partly based on Lehtonen and Veijanen (2018).
Original languageEnglish
JournalThe Survey Statistician
Volume79
Pages (from-to)16-26
Number of pages11
ISSN2521-991X
Publication statusPublished - Jan 2019
MoE publication typeB1 Journal article

Fields of Science

  • 112 Statistics and probability
  • model-free calibration
  • model-assisted calibration
  • mixed models
  • empirical best predictor
  • design-based simulation experiments

Cite this

@article{f67db7d3d36c4e9c884b1b739b28f74d,
title = "Calibration methods for small domain estimation",
abstract = "Abstract. Small area estimation was discussed in this section of The Survey Statistician in the July 2010 issue by Danny Pfeffermann in his article {"}Small Area Estimation: Basic Concepts, Models and Ongoing Research{"}. A more comprehensive review article was published a couple years later (Pfeffermann 2013). The Wiley book {"}Small Area Estimation, Second Edition{"} of 2015 by J.N.K. Rao and Isabel Molina presents an update to Rao's monograph of 2003 on small area estimation. These important sources cover model-based and design-based approaches on small area estimation (SAE) and show in particular the progress in model-based methods, and the progress is ongoing. In this article we introduce methods that incorporate assisting models in a design-based estimation procedure for population characteristics (totals, means etc.) for subgroups or domains, including small domains (with small sample size). We use logistic mixed models in model-assisted calibration estimation of poverty rates for administrative regions (domains of interest). Statistical properties (design bias and accuracy) of the method is compared with the classical model-free calibration method of Deville and S{\"a}rndal (1992) and further, with a model-based SAE method that relies on the same logistic mixed model as the model-assisted counterpart. Our design-based simulation experiments employ real data obtained from registers of Statistics Finland. The paper is partly based on Lehtonen and Veijanen (2018).",
keywords = "112 Statistics and probability, model-free calibration, model-assisted calibration, mixed models, empirical best predictor, design-based simulation experiments",
author = "Risto Lehtonen and Ari Veijanen",
year = "2019",
month = "1",
language = "English",
volume = "79",
pages = "16--26",
journal = "The Survey Statistician",
issn = "2521-991X",
publisher = "International Association of Survey Statisticians",

}

Calibration methods for small domain estimation. / Lehtonen, Risto; Veijanen, Ari.

In: The Survey Statistician, Vol. 79, 01.2019, p. 16-26.

Research output: Contribution to journalArticleScientific

TY - JOUR

T1 - Calibration methods for small domain estimation

AU - Lehtonen, Risto

AU - Veijanen, Ari

PY - 2019/1

Y1 - 2019/1

N2 - Abstract. Small area estimation was discussed in this section of The Survey Statistician in the July 2010 issue by Danny Pfeffermann in his article "Small Area Estimation: Basic Concepts, Models and Ongoing Research". A more comprehensive review article was published a couple years later (Pfeffermann 2013). The Wiley book "Small Area Estimation, Second Edition" of 2015 by J.N.K. Rao and Isabel Molina presents an update to Rao's monograph of 2003 on small area estimation. These important sources cover model-based and design-based approaches on small area estimation (SAE) and show in particular the progress in model-based methods, and the progress is ongoing. In this article we introduce methods that incorporate assisting models in a design-based estimation procedure for population characteristics (totals, means etc.) for subgroups or domains, including small domains (with small sample size). We use logistic mixed models in model-assisted calibration estimation of poverty rates for administrative regions (domains of interest). Statistical properties (design bias and accuracy) of the method is compared with the classical model-free calibration method of Deville and Särndal (1992) and further, with a model-based SAE method that relies on the same logistic mixed model as the model-assisted counterpart. Our design-based simulation experiments employ real data obtained from registers of Statistics Finland. The paper is partly based on Lehtonen and Veijanen (2018).

AB - Abstract. Small area estimation was discussed in this section of The Survey Statistician in the July 2010 issue by Danny Pfeffermann in his article "Small Area Estimation: Basic Concepts, Models and Ongoing Research". A more comprehensive review article was published a couple years later (Pfeffermann 2013). The Wiley book "Small Area Estimation, Second Edition" of 2015 by J.N.K. Rao and Isabel Molina presents an update to Rao's monograph of 2003 on small area estimation. These important sources cover model-based and design-based approaches on small area estimation (SAE) and show in particular the progress in model-based methods, and the progress is ongoing. In this article we introduce methods that incorporate assisting models in a design-based estimation procedure for population characteristics (totals, means etc.) for subgroups or domains, including small domains (with small sample size). We use logistic mixed models in model-assisted calibration estimation of poverty rates for administrative regions (domains of interest). Statistical properties (design bias and accuracy) of the method is compared with the classical model-free calibration method of Deville and Särndal (1992) and further, with a model-based SAE method that relies on the same logistic mixed model as the model-assisted counterpart. Our design-based simulation experiments employ real data obtained from registers of Statistics Finland. The paper is partly based on Lehtonen and Veijanen (2018).

KW - 112 Statistics and probability

KW - model-free calibration

KW - model-assisted calibration

KW - mixed models

KW - empirical best predictor

KW - design-based simulation experiments

M3 - Article

VL - 79

SP - 16

EP - 26

JO - The Survey Statistician

JF - The Survey Statistician

SN - 2521-991X

ER -