| Everything is related to everything else, but near things are more related than distant things. Spatial autocorrelation is a useful quantitative index which can reveal regularity in the spatial distribution. In this paper, a spatial autocorrelation analysis was undertaken to investigate the differences between Local Moran Index and Local Gi* based on the data of sex ratio of birth of Jiangsu province. Emphesis was focused on how to choose suitable indexes(weight, the numbers of neighbors, etc.) that could show the characteristics of spatial distribution in sex ratio of birth.Sex ratio of birth of Jiangsu province was as high as 116.73 in 2000. The high cluster of sex ratio of birth was found at the north of Jiangsu while the low cluster of sex ratio of birth was located at the south of Jiangsu. There were obvious regional differences. Such pattern revealed by spatial autocorrelation csn provide decision support for the government in dealing with the highly skewed sex ratio at birth.Data used in this research was collected from the 5th Census in 2000. With counties as the basic spatial units. This paper puts emphesis on the boundary effects, which was caused by whether to include the neighboring counties in other provinces. These neighbors (1st, 2nd and 3rd order neighbors) chosen for the analysis and the result was compared with the situation without these neighbors.ArcGIS and Geoda Software were used in this paper. The paper discussed issues as follows: the data of Jiangsu counties were processed in order to match the attribute data and vector data. Rook weight, K-nearest weight and fixed distance weight were used to calculate different weight matrix. Visual basci codes was written to implement format transformation of weight files that can be used in ArcGIS (for Gi* tools). At last, thematic maps were plotted for analysis. This paper conducted four comparisons :(1) Local Moran's I vs. local Gi*; (2) Including outside counties vs. not including outside counties; (3) Comparison of different weight matrix;(4) Comparison between 1st, 2nd and 3rd order Rook weight.Several conclusions were made as below:1. Each index showed its own advantage according to their outcomes of cluster figure. The high-high cluster of local Moran's I was equivalent to the hot spot of Gi* and the low-low cluster is equivalent to the cold spot of Gi*. The main difference lied on that the high-low and low-high cluster in local Moran's I reflected alienation in homogeneous area while the hot and cold spots of Gi* showed extension. The Gi* got higher spatial autocorrelation and wider range. These two methods could be combined for research.2. It was thought that 1st order Rook weight was probobaly the most suitable weight. Because 1st order Rook weight indicates the true physical adjacency. Sex ratio of birth may get more influence through adjacent regions, such as the spread of patriarchal ideology, the diffusion of prenatal sex identification technology.3. Spatial autocorrelation analysis was impacted by the concept of boundaryThe sex ratio at birth of Jiangsu was obviousely impacted by neighbors of other provinces. So, these neighboring counties should be considered in the analysis. The results will provide more useful reference for decision-makers..The calculator results showed that the high cluster of sex ratio of birth was located at the north of Jiangsu and the low cluster of sex ratio of birth was located at the south of Jiangsu. It could be many reasons: the economic underdeveloped in north Jiangsu may bring out thoughts and culture lagged behind, many neighbors of north counties of Jiangsu had high sex ratio birth also lead to high cluster.As sex ratio at birth was a social problem, which had strong spatial correlation. We believe that spatial autocorrelation analysis would have a bright prospect in reveal the spatial pattern of sex ratio at birth.
|
PDF Full Text Request