nhurst

Assigning Randomly Generated Clusters to Parcel Data

Discussion created by nhurst on Jul 15, 2012
Hi Everyone,

Is there a way to randomly cluster groups of parcel/property data together based on proximity?  I want to spatially join about 100,000 points (properties in Minneapolis) to 1KM square overlays (thereby assigning a unique ID to each cluster of parcels).  I'd like to do this 100 times over, with slight variations in the configuration of the clusters--like shifting the 1km grid over slightly each time.  So in the end, I'd have a data set (observations=properties) with 100 columns of uniquely identified clusters. 

Background:

I am working on a difference-in-difference OLS analysis of land use change and light-rail transit proximity in Minneapolis.  My primary software for data preparation is ArcGIS, but most of the analysis has been done in STATA.  I'd like to run one estimation for each cluster overlay, and then find the mean coefficients and the standard deviation of all the generated coefficients...

Essentially, this would be a kind of "bootstrapping," using GIS in the process.

Thoughts?

Thanks,
Needham

Outcomes