For my masters dissertation I'm looking at the socioeconomic impact of Edinburgh's historic trolley system. I'd like to do something with spatial autocorrelation and clustering along the network but have been all but completely unsuccessful.
I've managed to get the 'Cluster and Outlier Analysis (Anselin Local Moran's I)' tool to run twice in 3 days, and only once correctly.
I made a test file of points, coded them with rents and deliberately clustered or scattered them. I ran it with inverse distance and the one time it worked and gave me sensible answers. I haven't been able to get the program to run with the same settings again.
The process has also run once using a weights file that I created, using the methodology outlined here: <http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=Modeling%20spatial%20relationships>. The answers it gave me made no sense, however, with the LMicluster field showing very negative values (e.g. -84.3873846) for my clustered points and very positive value for my scattered points. I haven't been able to get the tool to run again.
I keep getting a popup that tells me that python.exe has encountered a problem and needs to close. When I searched for the specific error message online, I came upon this forum thread <http://forums.esri.com/Thread.asp?c=93&f=983&t=248033 >, but I know the department has the right version of Python for 9.2.
I'm out of ideas and so is the IT staff here. Can you help, or know where else to look?
Hi Angela, I'm really sorry you are having problems. I've not heard of that problem before, but just out of curiousity are you using Service Pack 6 for ArcGIS 9.2? Are you using projected data? You should always use projected data with the spatial statistics tools. The Python errors do suggest an installation issue... but usually the tools don't work at all in that case (you indicated you got them to work twice). We have seen problems when there are multiple versions of python installed. If you have multiple versions of python on the machine (python 2.1 and 2.4?), that can create problems. Here is the link to the article about this problem: http://resources.arcgis.com/content/kbase?fa=articleShow&d=31909 Can you please see if you are able to run the MultiRing Buffer tool? If this tool runs fine, it probably is not an installation issue.
Otherwise, if it might be possible for you to send me your data, I can set up a machine with software matching what you are using, and can try to reproduce the problem. Please feel free to contact me directly and I will do my best to help you with this: LScott@ESRI.com
Again, I'm sorry you are having problems with these tools. Lauren Scott ESRI Geoprocessing, Spatial Statistics
I do not have insight as to why the crash is happening but it sounds like you may be violating assumptions of point pattern statistics (PPA). PPA assumes both a homogeneous point process and isotropy. Point patterns along a linear dependency (network) violate both of these assumptions. Additionally, most PPA statistics compare against an expected complete spatial random (CSR) process. Expected CSR is derived from a know per-area unit of spatial distribution given the extent of the study area. Since observations along networks are linearly dependent the expected CSR is not correct. There are a few papers out that have modified the Ripley's K and KDE's for linear features, but these modifications are not available in the spatial statistics ArcToolbox. Okabe has developed an ArcGIS Toolbox with a variety of network PPA tools (http://sanet.csis.u-tokyo.ac.jp/index.html) but I am not sure if it is free anymore.
Okabe, A. and Satoh, T. (2009) Spatial analysis on a network, in Fotheringham, A.S. and Rogers, P.A. (eds.) The SAGE Handbook on Spatial Analysis, 443-464, London: SAGE
Okabe, A., Okunuki, K. and Shiode, S. (2006) SANET: A toolbox for spatial analysis on a network. Geographical Analysis, 38(1), 57-66
Shiode, S. (2008) Analysis of a distribution of point events using the network-based quadrat method, Geographical Analysis 40(4), 401-422.
