Model-based small area estimation at two scales using Moran's spatial filtering

Phuong N. Truong*, Alfred Stein

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

In spatial epidemiology and public health studies, including covariates in small area estimation of spatial binary data remains a challenge. In this paper, Moran's spatial filtering is proposed to model two-scale spatial binary data. Two models are developed: the first uses deterministic estimation of the sample size at small areal level; the second generates a random sample size using the multinomial distribution. The models were applied to estimate the underweight among children at Vietnamese district level using sampling survey data at provincial level. The results show that the first model outperformed the second model regarding its accuracy and simplicity. Eigenvector maps improve model parameter estimation, and allow for the effects of spatial spillover and covariates. Prediction at the district level indicates that many underweight children came from the mountainous areas in 2014. The study concludes that the proposed models serve as alternatives to small area estimation of spatial binary data.

Original languageEnglish
Article number100303
JournalSpatial and Spatio-temporal Epidemiology
Volume31
DOIs
Publication statusPublished - 1 Nov 2019

    Fingerprint

Keywords

  • Childhood malnutrition
  • Moran's spatial filtering
  • Sample survey data
  • Small area estimation
  • Spatial binary data

Cite this