Application of sandwich spatial estimation method in cancer mapping: A case study for breast cancer mortality in the Chinese mainland, 2005. (December 2019)

• Record Type: Journal Article
• Authors: Liao, Yilan; Li, Dongyue; Zhang, Ningxu; Xia, Changfa; Zheng, Rongshou; Zeng, Hongmei; Zhang, Siwei; Wang, Jinfeng; Chen, Wanqing
• Is Part Of: Statistical methods in medical research. Volume 28:Number 12(2019)
• Page Start: 3609
• Page End: 3626
• Publication Date: 2019-12
• Subjects: Sandwich estimation -- Bayesian -- cancer mortality -- spatial distribution -- mapping; Medicine -- Research -- Statistical methods -- Periodicals; Research -- Periodicals; Review Literature -- Periodicals; Statistics -- methods -- Periodicals; Médecine -- Recherche -- Méthodes statistiques -- Périodiques; 610.727
• DOI: 10.1177/0962280218811344
• Languages: English
• ISSNs: 0962-2802

Abstract

High-accuracy spatial distribution estimation is crucial for cancer prevention and control. Due to their complicated pathogenic factors, the distributions of many cancers' mortalities appear blocky, and spatial heterogeneity is common. However, most of the commonly used cancer mapping methods are based on spatial autocorrelation theory. Sandwich estimation is a new method based on spatial heterogeneity theory. A modified sandwich estimation method suitable for the estimation of cancer mortality distribution is proposed in this study. The variances of cancer mortality data are used to fuse sandwich estimation results from various auxiliary variables, the feasibility of which in estimating cancer mortality distributions is explained theoretically. The breast cancer (BC) mortality of the Chinese mainland in 2005 was taken as a case, and the accuracy of the modified sandwich estimation method was compared with that of the Hierarchical Bayesian (HB), the Co-Kriging (CK) and the Ordinary Kriging (OK) methods. The accuracy of the modified sandwich estimation method was better than the HB, the CK and the OK methods, and the estimation result from the modified sandwich estimation method was more likely to be acceptable. Therefore, this study represents an attempt to apply the sandwich estimation method to the estimation of cancer mortality distributions with strong spatial heterogeneity, which holds great potential for further application.