MSA 8901 Project

Spatial Autocorrelation and Rasterization Error –

A determination of accuracy of spatial statistical inference on spatial data converted from vector to raster data model


Abstract.  When performing conversion from vector data model to raster data model on spatial data, rasterization error occurs.  This project is going to determine the degree of rasterization error as illustrated by the change of the degree of spatial autocorrelation as different raster grid cell sizes are used.  The data chosen is from 1996 census data on average annual household income in each enumeration area of Vancouver, BC.  A portion of Vancouver is selected for analysis with an area of 2500m by 2500m.  Two spatial statistics on autocorrelation widely used are the Moran’s I coefficient and the Geary’s c coefficient.  Both of them are calculated in S-Plus and the results are plotted.  The expected result of these two statistics is towards negative autocorrelation as grid cell size increases.  As expected, both statistics confirm this but after a threshold both fluctuate.  This suggests that determination of this threshold is important as to the accuracy of statistical inference when doing spatial analysis on raster dataset.


Content:

1. The Problem

2. Data and Method

3. Results and Outputs

4. Analysis of Results

5. Conclusion

6. References


This project is done by Christopher Au-Yeung, MSA student at the Joint Ryerson Polytechnic University and University of Toronto Master of Spatial Analysis Program. Summited to Dr. Ferko Csillag for MSA901 Accuracy of Spatial Databases.

Comments and Questions: cauyeung@sfu.ca