Exploring scale-dependent correlations between cancer mortality rates using factorial kriging and population-weighted semivariograms.

摘要: This paper presents a geostatistical methodology which accounts for spatially varying population size in the processing of cancer mortality data. The approach proceeds two steps: (1) spatial patterns are first described and modeled using population-weighted semivariogram estimators, (2) components corresponding to nested structures identified on semivariograms then estimated mapped variant factorial kriging. main benefit over traditional smoothers is that pattern variability (i.e. direction-dependent variability, range correlation, presence scales variability) directly incorporated into computation weights assigned surrounding observations. Moreover, besides filtering noise data procedure allows decomposition structured component several local versus regional basis models. A simulation study demonstrates maps closer underlying risk terms prediction errors provide better visualization than original rates or smoothed weighted linear averages. proposed also attenuates underestimation magnitude correlation between various resulting from attached has great potential explore scale-dependent risks developing cancers detect clusters at scales, should lead more accurate representation geographic variation risk, ultimately understanding causative relationships.