GIS 5990 Module 7: Scythian Mounds Report Week, Regression Analysis

Results of ordinary least squares regression analysis on Scythian mound data.

The third and final week of our work with Scythian mounds near Tuekta, Russia allowed us to dig deeper into statistical analyses. I am glad to get more practice with statistics; perhaps eventually I will come to understand statistics and how to apply statistical tests in practical situations. Until then, the detailed guidance in our assignment instructions on interpreting statistical test results was most welcome.

We began the assignment by creating 100 random points as examples of locations without sites; these were then merged with our point shapefile digitized previously representing mound locations. We then needed to include the category of elevation, slope, and aspect in which each point was located in the merged shapefile's attribute table. I used the Extract Multi Values to Points tool, as described in the instructions. This was much simpler than manually entering the data, although it did entail deleting superfluous fields in the attribute table.

After the data were prepared properly, we ran the statistical tools of Ordinary Least Squares Regression, Spatial Autocorrelation, and Hot Spot Analysis. The results of these analyses indicate all three environmental variables affect mound placement (with elevation being the most significant) and give a 99% confidence value that mound placement is not random. The detailed results of the Spatial Autocorrelation analysis is included in the above map.

This module taught me much on the proper way to set up regression analysis. I will likely use this module for reference in my internship and future work until I become more comfortable with statistical work.