期刊名称：American Journal of Geographic Information System
印刷版ISSN：2163-1131
电子版ISSN：2163-114X
出版年度：2015
卷号：4
期号：3
页码：87-94
DOI：10.5923/j.ajgis.20150403.01
语种：English
出版社：Scientific & Academic Publishing Co.
摘要：This paper presents comparative analytical procedures of spatial autocorrelation indices for identifying cluster and modelling spatial pattern of tuberculosis (TB) prevalence. It compares global and Local Indicators of Spatial Association for finding cluster and hotspot locations. It also assessed three interpolation methods for disaggregating TB prevalence to enable analysis of local cluster and hotspot locations due to data limitation. Inverse Distance Weighting (IDW) was selected for its lowest variance (3.4) and Root Mean Square Error (RMSE) in comparison to other models. The global Moran’ I test was -0.09 and a p-value of 0.70 using aggregated TB prevalence for the year 2008 as the best case scenario considering all the years. Hence, global Moran’ I do not identify any significant cluster. In the case of general G, the situation is different for the clustering pattern, though quite low in some years but even unique to year 2010 with significant high z value (1.14) and very low p-value (0.25). When predicted data were used for these indices, both measures have shown interesting results with Moran’s I index of 0.86 which indicate high cluster of TB prevalence in the area. The analysis of Local Indicators of Spatial Association which were carried out based on the most accurate interpolated (IDW) values indicate high and low clustered areas, hotspot and cold spot locations. Despite data limitations, the outcome of this study is still very useful in assigning measures to minimize TB prevalence and for changing policy options in health service provision.
