关键词: 尺度矩阵, 空间自相关, 扫描统计量, 围产儿, 聚集, 甘肃

Abstract:

The problem of aggregation has been always one of the hot and focuses of the region research, and the positioning of aggregation is a prerequisite for continuing in-depth study in aggregation analysis. Aggregation appears a high degree of sensitivity on the scale. When the aggregation was analyzed by using the spatial autocorrelation method, scale choices are often susceptible to subjective judgments of the researchers, and exist the possibility of selection bias. So the spatial weight problem has been controversial. In addition, the aggregation is obviously space-dependent as well as time-dependent since different aggregation takes place within different time and space, which is neglected by the spatial autocorrelation method. Therefore, some scholars have been trying to explore better ways for aggregation analysis. In comparison, the spatial-temporal scan statistic method raised by scholars as Kulldorff shows its superiority. This paper targets at 51 districts and counties within the central and southern regions of Gansu Province, which are densely populated minority areas. Using software Geoda0.9.5 and Clusterseer0.2.3, spatial autocorrelation analysis and space time scan statistical analysis were adopted about 30463 perinatal deaths during 2001 to 2010 according to the report delivered by monitoring points of Ministry of Health. It also gives a detailed comparison between spatial autocorrelation and scanning statistic from the different perspectives of scale choices, scale transforming and space integration, and then tests the analysis results by appealing Sam 4.0. Through theoretical and empirical analysis, it further confirms that spatial-temporal scan statistic method has the significant advantage from the three aspects of scale selection, scale conversion, and spatial-temporal integration,which shows that it is not only effective measure to solve the problem of artificial selection, but also to achieve the scale extrapolation and automatic conversion, and more favorable mix of three-dimensional, dynamic, multi-scale analysis. Therefore, a conclusion is drawn that the space-time scan statistic method is superior than other methods. Specifically, there are three main aspects: 1） Compared to spatial autocorrelation methods and space-time scan statistic method, the adjacency and distance matrix is artificial selection in spatial autocorrelation. The space-time scan statistic method achieves the purpose of scale automatically conversion and avoids the instability. 2) The scale of spatial autocorrelation methods is static, single, and the scale of space-time scan statistic is dynamic, three-dimensional and multi-scale.3) In analyzing the issue of spatial autocorrelation method does not take into account the time factor, while the space-time scan statistic method is not only the full integration of space technology, and achieves good analytical results. By combining the time factor, the range of aggregation can be reduced to a more accurate range. Moreover, space-time scan statistic method can even make a prediction, which can provide a basis for decision-making.

Key words: scale matrix, spatial autocorrelation, spatial-temporal proximity, scan statistic, perinatal deaths, aggregation, Gansu